<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Field Guide · Brennenstuhl on Security</title><link>https://www.janbrennenstuhl.eu/knowledge-base/</link><description>Short reference entries on identity, authentication and account security. Each one answers a question with a claim, shows the evidence behind it and states its review status.</description><language>en-GB</language><copyright>Original writing CC BY 4.0, Jan Brennenstuhl</copyright><atom:link href="https://www.janbrennenstuhl.eu/knowledge-base/index.xml" rel="self" type="application/rss+xml"/><item><title>Should an agent use a robot account or delegated access?</title><link>https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/delegation-vs-robot-accounts/</link><guid isPermaLink="true">https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/delegation-vs-robot-accounts/</guid><pubDate>Tue, 13 Oct 2026 00:00:00 +0200</pubDate><description>&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; When an agent acts for a person, give it delegated access that names both the person and the agent, with narrow scopes, an expiry, and a way to revoke it. A robot account shared by several people gives the agent their combined permissions and hides whose request was whose.&lt;/p&gt;&lt;h2 id="key-findings"&gt;Key findings &lt;a class="heading-anchor" href="#key-findings" aria-label="Link to section: Key findings"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;figure class="finding" id="f-agent-robot-account-combined-permissions"&gt;
&lt;blockquote&gt;&lt;p&gt;When several people share an agent that holds a robot account, the agent acts with the combined permissions of all its users, and nobody can tell whose request was whose.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.janbrennenstuhl.eu/agentic-identity-broker/" class="ext"&gt;&lt;cite&gt;What I Learned Building an Identity Broker With Coding Agents (janbrennenstuhl.eu, 2026)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-agent-robot-account-combined-permissions" title="#f-agent-robot-account-combined-permissions"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="When several people share an agent that holds a robot account, the agent acts with the combined permissions of all its users, and nobody can tell whose request was whose." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/delegation-vs-robot-accounts/#f-agent-robot-account-combined-permissions" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/delegation-vs-robot-accounts/f-agent-robot-account-combined-permissions.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-agent-delegation-keeps-actor-identity"&gt;
&lt;blockquote&gt;&lt;p&gt;Under the delegation semantics of OAuth 2.0 Token Exchange, the acting party keeps its own identity, separate from the person whose rights it uses.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.rfc-editor.org/rfc/rfc8693#section-1.1" class="ext"&gt;&lt;cite&gt;RFC 8693, OAuth 2.0 Token Exchange (IETF, 2020), Section 1.1&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-agent-delegation-keeps-actor-identity" title="#f-agent-delegation-keeps-actor-identity"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Under the delegation semantics of OAuth 2.0 Token Exchange, the acting party keeps its own identity, separate from the person whose rights it uses." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/delegation-vs-robot-accounts/#f-agent-delegation-keeps-actor-identity" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/delegation-vs-robot-accounts/f-agent-delegation-keeps-actor-identity.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-agent-token-least-privilege"&gt;
&lt;blockquote&gt;&lt;p&gt;The OAuth 2.0 security best current practice says an access token&amp;rsquo;s privileges should be restricted to the minimum the use case requires, and the token should be audience-restricted to a specific resource server.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.rfc-editor.org/rfc/rfc9700#section-2.3" class="ext"&gt;&lt;cite&gt;RFC 9700, Best Current Practice for OAuth 2.0 Security (IETF, 2025), Section 2.3&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-agent-token-least-privilege" title="#f-agent-token-least-privilege"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="The OAuth 2.0 security best current practice says an access token&amp;rsquo;s privileges should be restricted to the minimum the use case requires, and the token should be audience-restricted to a specific resource server." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/delegation-vs-robot-accounts/#f-agent-token-least-privilege" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/delegation-vs-robot-accounts/f-agent-token-least-privilege.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-agent-no-long-lived-keys"&gt;
&lt;blockquote&gt;&lt;p&gt;Coding agents should not hold persistent identity and access management (IAM) roles or long-lived API keys; an identity broker can issue short-lived, task-scoped credentials on demand instead.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.janbrennenstuhl.eu/appsec-agentic-engineering/" class="ext"&gt;&lt;cite&gt;AppSec in the Age of Agentic Engineering (janbrennenstuhl.eu, 2026)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-agent-no-long-lived-keys" title="#f-agent-no-long-lived-keys"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Coding agents should not hold persistent identity and access management (IAM) roles or long-lived API keys; an identity broker can issue short-lived, task-scoped credentials on demand instead." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/delegation-vs-robot-accounts/#f-agent-no-long-lived-keys" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/delegation-vs-robot-accounts/f-agent-no-long-lived-keys.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;h2 id="evidence"&gt;Evidence &lt;a class="heading-anchor" href="#evidence" aria-label="Link to section: Evidence"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;div class="table-scroll studies" tabindex="0" role="region" aria-label="Primary sources on robot accounts and delegated agent access"&gt;&lt;table&gt;
&lt;caption&gt;Primary sources on robot accounts and delegated agent access&lt;/caption&gt;
&lt;thead&gt;&lt;tr&gt;&lt;th scope="col"&gt;Study&lt;/th&gt;&lt;th scope="col"&gt;Year&lt;/th&gt;&lt;th scope="col"&gt;Sample&lt;/th&gt;&lt;th scope="col"&gt;Method&lt;/th&gt;&lt;th scope="col"&gt;Finding&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.rfc-editor.org/rfc/rfc8693" class="ext"&gt;OAuth 2.0 Token Exchange (RFC 8693)&lt;/a&gt;&lt;span class="org"&gt;IETF&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2020&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Specification&lt;/td&gt;&lt;td&gt;Defines delegation and impersonation semantics and the act claim that names the acting party.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.rfc-editor.org/rfc/rfc9700" class="ext"&gt;Best Current Practice for OAuth 2.0 Security (RFC 9700)&lt;/a&gt;&lt;span class="org"&gt;IETF&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2025&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Specification&lt;/td&gt;&lt;td&gt;Asks for least-privilege, audience-restricted tokens and warns when resource servers can&amp;rsquo;t tell client tokens from user tokens.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://modelcontextprotocol.io/specification/2026-07-28/basic/authorization" class="ext"&gt;Model Context Protocol: Authorization (revision 2026-07-28)&lt;/a&gt;&lt;span class="org"&gt;Model Context Protocol&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2026&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Specification&lt;/td&gt;&lt;td&gt;Treats clients acting for a user differently from clients acting on their own behalf.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.janbrennenstuhl.eu/agentic-identity-broker/" class="ext"&gt;What I Learned Building an Identity Broker With Coding Agents&lt;/a&gt;&lt;span class="org"&gt;this site&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2026&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Field report&lt;/td&gt;&lt;td&gt;Explains why the Agentic Identity Broker keeps both the user and the agent in every authorization decision.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.janbrennenstuhl.eu/appsec-agentic-engineering/" class="ext"&gt;AppSec in the Age of Agentic Engineering&lt;/a&gt;&lt;span class="org"&gt;this site&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2026&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Essay&lt;/td&gt;&lt;td&gt;Argues for short-lived, task-scoped credentials and user-revocable delegation for agents.&lt;/td&gt;&lt;/tr&gt;
&lt;/tbody&gt;&lt;/table&gt;&lt;/div&gt;
&lt;h2 id="what-to-do"&gt;What to do &lt;a class="heading-anchor" href="#what-to-do" aria-label="Link to section: What to do"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="for-engineers"&gt;For engineers &lt;a class="heading-anchor" href="#for-engineers" aria-label="Link to section: For engineers"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;Ask first whom the agent acts for. If it runs a job on its own behalf, a dedicated client with its own narrow permissions is a reasonable fit. The &lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/mcp-authorization/"&gt;MCP authorization specification&lt;/a&gt; explicitly mentions such &lt;code&gt;client_credentials&lt;/code&gt; clients. If it acts for a person, use a grant that names that person and that agent, and keep it narrow and short-lived as &lt;a href="https://www.rfc-editor.org/rfc/rfc9700#section-2.3" class="ext"&gt;RFC 9700&lt;/a&gt; asks.&lt;/p&gt;
&lt;p&gt;Prefer delegation over impersonation. A delegated token can carry the agent as the actor next to the person as the subject, so a resource server and your logs can tell them apart. RFC 9700 also warns that a resource server that can&amp;rsquo;t distinguish tokens obtained with a person&amp;rsquo;s involvement from tokens a client obtained on its own may let that client reach the person&amp;rsquo;s resources.&lt;/p&gt;
&lt;p&gt;Keep provider credentials out of the agent&amp;rsquo;s context. &lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/token-exchange/"&gt;Token exchange&lt;/a&gt; shows how a gateway can do this on each tool call.&lt;/p&gt;
&lt;h3 id="for-product-and-ux"&gt;For product and UX &lt;a class="heading-anchor" href="#for-product-and-ux" aria-label="Link to section: For product and UX"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;Show people which agent holds access to which service, with what permissions and until when. Give them one place to &lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/consent-and-revocation/"&gt;revoke that access&lt;/a&gt;. If several people use the same agent, show each of them only the grants they made.&lt;/p&gt;
&lt;h2 id="how-it-works"&gt;How it works &lt;a class="heading-anchor" href="#how-it-works" aria-label="Link to section: How it works"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;p&gt;With a robot account, the agent holds a long-lived credential for each MCP server or API it calls. In OAuth terms, that is closest to a client that gets tokens for itself, without a person in the flow. RFC 9700 quotes the JSON Web Token (JWT) profile for access tokens (RFC 9068): the &lt;code&gt;sub&lt;/code&gt; claim of such a token should identify the client application, while a token obtained with a person&amp;rsquo;s involvement should identify that person. A robot account therefore can&amp;rsquo;t express who asked for an action. In &lt;a href="https://www.janbrennenstuhl.eu/agentic-identity-broker/"&gt;my retrospective&lt;/a&gt;, I describe the consequences once several people share the agent: combined permissions, no attribution, and credentials within reach of anything that tricks the agent.&lt;/p&gt;
&lt;p&gt;&lt;a href="https://www.rfc-editor.org/rfc/rfc8693#section-1.1" class="ext"&gt;RFC 8693&lt;/a&gt; describes 2 ways one party can use another&amp;rsquo;s rights. With impersonation, party A gets B&amp;rsquo;s rights and is indistinguishable from B in that context. With delegation, A keeps its own identity, and it&amp;rsquo;s explicitly understood that A acts for B. The RFC uses this exact phrase: &amp;ldquo;In a sense, A is an agent for B.&amp;rdquo; A token that carries both identities is a composite token. In a JWT, the &lt;code&gt;act&lt;/code&gt; claim names the acting party.&lt;/p&gt;
&lt;p&gt;The Agentic Identity Broker that Magnus Jungsbluth and I built applies delegation. A user grants an agent access to a service with a chosen permission set and an optional expiry and can revoke the grant at any time. The broker keeps both the user and the agent in every authorization decision. The essay and our &lt;a href="https://www.janbrennenstuhl.eu/talks/economies-of-scale-for-mcp/"&gt;AGNTCon + MCPCon talk&lt;/a&gt; explain the design, and the &lt;a href="https://www.janbrennenstuhl.eu/projects/agentic-identity-broker/"&gt;project page&lt;/a&gt; links the code.&lt;/p&gt;
&lt;p&gt;This model covers agents that act for a person. For autonomous agents with no human in the loop, I wrote in the retrospective that every approach I know of is still brittle.&lt;/p&gt;
&lt;h2 id="related"&gt;Related &lt;a class="heading-anchor" href="#related" aria-label="Link to section: Related"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/token-exchange/"&gt;How does OAuth 2.0 Token Exchange keep provider tokens away from agents?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/consent-and-revocation/"&gt;How should people grant and revoke an agent&amp;rsquo;s access?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/mcp-authorization/"&gt;How does authorization work for MCP servers?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/appsec-agentic-engineering/"&gt;AppSec in the Age of Agentic Engineering&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>What should a password policy require?</title><link>https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-policy-requirements/</link><guid isPermaLink="true">https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-policy-requirements/</guid><pubDate>Tue, 13 Oct 2026 00:00:00 +0200</pubDate><description>&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; Under NIST SP 800-63B-4, a password used on its own must have at least 15 characters, or 8 when it is one factor of multi-factor authentication, and systems should accept at least 64. Composition rules, forced periodic changes, hints and security questions are out; a blocklist of common and breached passwords is in.&lt;/p&gt;&lt;h2 id="key-findings"&gt;Key findings &lt;a class="heading-anchor" href="#key-findings" aria-label="Link to section: Key findings"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;figure class="finding" id="f-nist-min-length-15"&gt;
&lt;blockquote&gt;&lt;p&gt;NIST SP 800-63B-4 requires passwords used as the only authentication factor to be at least 15 characters long, and at least 8 when the password is used only as part of multi-factor authentication.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://pages.nist.gov/800-63-4/sp800-63b/authenticators/" class="ext"&gt;&lt;cite&gt;NIST SP 800-63B-4, Section 3.1.1.2&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-nist-min-length-15" title="#f-nist-min-length-15"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="NIST SP 800-63B-4 requires passwords used as the only authentication factor to be at least 15 characters long, and at least 8 when the password is used only as part of multi-factor authentication." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-policy-requirements/#f-nist-min-length-15" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-policy-requirements/f-nist-min-length-15.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-nist-no-composition-no-rotation"&gt;
&lt;blockquote&gt;&lt;p&gt;Verifiers shall not impose composition rules such as mixing character types, and shall not require periodic password changes, but shall force a change when there is evidence of compromise.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://pages.nist.gov/800-63-4/sp800-63b/authenticators/" class="ext"&gt;&lt;cite&gt;NIST SP 800-63B-4, Section 3.1.1.2&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-nist-no-composition-no-rotation" title="#f-nist-no-composition-no-rotation"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Verifiers shall not impose composition rules such as mixing character types, and shall not require periodic password changes, but shall force a change when there is evidence of compromise." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-policy-requirements/#f-nist-no-composition-no-rotation" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-policy-requirements/f-nist-no-composition-no-rotation.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-nist-blocklist-with-reason"&gt;
&lt;blockquote&gt;&lt;p&gt;A new password shall be compared in full against a blocklist of commonly used, expected or compromised passwords, and a rejected user shall be told why.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://pages.nist.gov/800-63-4/sp800-63b/authenticators/" class="ext"&gt;&lt;cite&gt;NIST SP 800-63B-4, Section 3.1.1.2&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-nist-blocklist-with-reason" title="#f-nist-blocklist-with-reason"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="A new password shall be compared in full against a blocklist of commonly used, expected or compromised passwords, and a rejected user shall be told why." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-policy-requirements/#f-nist-blocklist-with-reason" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-policy-requirements/f-nist-blocklist-with-reason.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-nist-password-managers"&gt;
&lt;blockquote&gt;&lt;p&gt;Verifiers shall allow password managers and autofill, should permit paste, and should accept passwords of at least 64 characters.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://pages.nist.gov/800-63-4/sp800-63b/authenticators/" class="ext"&gt;&lt;cite&gt;NIST SP 800-63B-4, Section 3.1.1.2&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-nist-password-managers" title="#f-nist-password-managers"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Verifiers shall allow password managers and autofill, should permit paste, and should accept passwords of at least 64 characters." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-policy-requirements/#f-nist-password-managers" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-policy-requirements/f-nist-password-managers.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;h2 id="evidence"&gt;Evidence &lt;a class="heading-anchor" href="#evidence" aria-label="Link to section: Evidence"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;div class="table-scroll studies" tabindex="0" role="region" aria-label="Sources"&gt;&lt;table&gt;
&lt;caption&gt;Sources&lt;/caption&gt;
&lt;thead&gt;&lt;tr&gt;&lt;th scope="col"&gt;Study&lt;/th&gt;&lt;th scope="col"&gt;Year&lt;/th&gt;&lt;th scope="col"&gt;Sample&lt;/th&gt;&lt;th scope="col"&gt;Method&lt;/th&gt;&lt;th scope="col"&gt;Finding&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://pages.nist.gov/800-63-4/sp800-63b/authenticators/" class="ext"&gt;NIST SP 800-63B-4: Digital Identity Guidelines, Authentication and Authenticator Management&lt;/a&gt;&lt;span class="org"&gt;NIST&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2025&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Standard&lt;/td&gt;&lt;td&gt;Section 3.1.1.2 sets the password verifier requirements summarized here.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://pages.nist.gov/800-63-4/sp800-63b/passwords/" class="ext"&gt;SP 800-63B-4 Appendix A: Strength of Passwords&lt;/a&gt;&lt;span class="org"&gt;NIST&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2025&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Informative appendix&lt;/td&gt;&lt;td&gt;Explains why length and blocklists matter more than composition rules.&lt;/td&gt;&lt;/tr&gt;
&lt;/tbody&gt;&lt;/table&gt;&lt;/div&gt;
&lt;h2 id="what-to-do"&gt;What to do &lt;a class="heading-anchor" href="#what-to-do" aria-label="Link to section: What to do"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="for-engineers"&gt;For engineers &lt;a class="heading-anchor" href="#for-engineers" aria-label="Link to section: For engineers"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;ul&gt;
&lt;li&gt;Set the minimum by how the password is used: 15 characters where it is the only factor, 8 only where a second factor is always required.&lt;/li&gt;
&lt;li&gt;Accept long passwords and verify them in full; the essay on &lt;a href="https://www.janbrennenstuhl.eu/long-password-support/"&gt;long password support&lt;/a&gt; covers hashing limits and denial-of-service concerns.&lt;/li&gt;
&lt;li&gt;Replace composition rules with a blocklist check against breached and common passwords. The field-guide entry on &lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/credential-stuffing/"&gt;stolen credentials&lt;/a&gt; covers privacy-preserving breach checks.&lt;/li&gt;
&lt;li&gt;Make forms work with password managers: the right &lt;code&gt;autocomplete&lt;/code&gt; values and no paste blocking. See &lt;a href="https://www.janbrennenstuhl.eu/password-manager-friendly/"&gt;password-manager friendly login forms&lt;/a&gt;.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="for-product-and-ux"&gt;For product and UX &lt;a class="heading-anchor" href="#for-product-and-ux" aria-label="Link to section: For product and UX"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;ul&gt;
&lt;li&gt;Explain a rejection in plain words (&amp;ldquo;this password appears in known breaches&amp;rdquo;) instead of listing rules after the fact. NIST requires a reason.&lt;/li&gt;
&lt;li&gt;Drop scheduled password expiry. Forced changes are for evidence of compromise, which makes them rarer and easier to explain.&lt;/li&gt;
&lt;li&gt;My &lt;a href="https://www.janbrennenstuhl.eu/password-policy/"&gt;2023 password policy essay&lt;/a&gt; recommended at least 12 characters. For passwords used on their own, the final SP 800-63B-4 now sets 15.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="check-your-policy"&gt;Check your policy &lt;a class="heading-anchor" href="#check-your-policy" aria-label="Link to section: Check your policy"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;section class="explainer" data-explainer="password-policy" aria-labelledby="ex-password-policy"&gt;
&lt;h3 id="ex-password-policy"&gt;Compare a policy with SP 800-63B-4&lt;/h3&gt;
&lt;div class="explainer-static"&gt;&lt;p&gt;The checker on this page compares the settings of a password policy with the requirements below. Without JavaScript, use the list as a checklist. &amp;ldquo;Shall&amp;rdquo; is a requirement, &amp;ldquo;should&amp;rdquo; a recommendation.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Shall:&lt;/strong&gt; at least 15 characters for single-factor passwords; at least 8 if the password is only one factor of MFA.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Should:&lt;/strong&gt; accept at least 64 characters, all printing ASCII characters, the space character and Unicode; count each Unicode code point as one character.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Shall not:&lt;/strong&gt; impose composition rules, such as requiring digits, symbols or mixed case.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Shall not:&lt;/strong&gt; require periodic password changes. &lt;strong&gt;Shall:&lt;/strong&gt; force a change when there is evidence of compromise.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Shall not:&lt;/strong&gt; offer password hints to unauthenticated users, or prompt for security questions when a password is chosen.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Shall:&lt;/strong&gt; verify the whole password, never a truncated or partial one.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Shall:&lt;/strong&gt; check new passwords against a blocklist, and tell the user why a password was rejected.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Shall:&lt;/strong&gt; rate-limit failed authentication attempts.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Shall:&lt;/strong&gt; allow password managers and autofill. &lt;strong&gt;Should:&lt;/strong&gt; allow paste and offer to show the password.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Shall:&lt;/strong&gt; store passwords salted and hashed with a suitable password hashing scheme, with a salt of at least 32 bits.&lt;/li&gt;
&lt;/ul&gt;&lt;/div&gt;
&lt;div class="explainer-app" hidden&gt;&lt;/div&gt;
&lt;/section&gt;
&lt;h2 id="related"&gt;Related &lt;a class="heading-anchor" href="#related" aria-label="Link to section: Related"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-strength-meter/"&gt;Do password strength meters work?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-reuse/"&gt;How common is password reuse?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/max-password-length/"&gt;Max password length: a secure password guide&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>What is inside a JWT?</title><link>https://www.janbrennenstuhl.eu/knowledge-base/oidc/jwt-anatomy/</link><guid isPermaLink="true">https://www.janbrennenstuhl.eu/knowledge-base/oidc/jwt-anatomy/</guid><pubDate>Tue, 13 Oct 2026 00:00:00 +0200</pubDate><description>&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; A signed JWT is 3 base64url-encoded parts joined by dots: a header that names the algorithm, a payload of claims, and a signature over both. The payload is encoded, not encrypted, so anyone holding the token can read it; only a signature check with a key you trust tells you whether to believe it.&lt;/p&gt;&lt;h2 id="key-findings"&gt;Key findings &lt;a class="heading-anchor" href="#key-findings" aria-label="Link to section: Key findings"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;figure class="finding" id="f-jwt-claims-in-jws-or-jwe"&gt;
&lt;blockquote&gt;&lt;p&gt;A JWT is a set of claims encoded as a JSON object inside a JWS or a JWE, so the same token format can be signed, encrypted, or both.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.rfc-editor.org/rfc/rfc7519#section-2" class="ext"&gt;&lt;cite&gt;RFC 7519, JSON Web Token (JWT), Section 2&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-jwt-claims-in-jws-or-jwe" title="#f-jwt-claims-in-jws-or-jwe"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="A JWT is a set of claims encoded as a JSON object inside a JWS or a JWE, so the same token format can be signed, encrypted, or both." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/oidc/jwt-anatomy/#f-jwt-claims-in-jws-or-jwe" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/oidc/jwt-anatomy/f-jwt-claims-in-jws-or-jwe.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-jwt-payload-is-readable"&gt;
&lt;blockquote&gt;&lt;p&gt;The payload of a signed JWT is only base64url-encoded, so a token that carries privacy-sensitive claims needs encryption or must not reach parties who should not read it.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.rfc-editor.org/rfc/rfc7519#section-12" class="ext"&gt;&lt;cite&gt;RFC 7519, Section 12, Privacy Considerations&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-jwt-payload-is-readable" title="#f-jwt-payload-is-readable"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="The payload of a signed JWT is only base64url-encoded, so a token that carries privacy-sensitive claims needs encryption or must not reach parties who should not read it." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/oidc/jwt-anatomy/#f-jwt-payload-is-readable" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/oidc/jwt-anatomy/f-jwt-payload-is-readable.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-jwt-registered-claims-optional"&gt;
&lt;blockquote&gt;&lt;p&gt;RFC 7519 registers 7 claim names, iss, sub, aud, exp, nbf, iat and jti, and none of them is mandatory in every case; each application profile decides which it requires.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.rfc-editor.org/rfc/rfc7519#section-4.1" class="ext"&gt;&lt;cite&gt;RFC 7519, Section 4.1&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-jwt-registered-claims-optional" title="#f-jwt-registered-claims-optional"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="RFC 7519 registers 7 claim names, iss, sub, aud, exp, nbf, iat and jti, and none of them is mandatory in every case; each application profile decides which it requires." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/oidc/jwt-anatomy/#f-jwt-registered-claims-optional" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/oidc/jwt-anatomy/f-jwt-registered-claims-optional.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-jwt-allowlist-algorithms"&gt;
&lt;blockquote&gt;&lt;p&gt;JWT libraries must let the caller specify the supported algorithms and must not use any other algorithm, which closes the door on tokens whose header was changed to &amp;ldquo;none&amp;rdquo;.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.rfc-editor.org/rfc/rfc8725#section-3.1" class="ext"&gt;&lt;cite&gt;RFC 8725, JSON Web Token Best Current Practices, Section 3.1&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-jwt-allowlist-algorithms" title="#f-jwt-allowlist-algorithms"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="JWT libraries must let the caller specify the supported algorithms and must not use any other algorithm, which closes the door on tokens whose header was changed to &amp;ldquo;none&amp;rdquo;." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/oidc/jwt-anatomy/#f-jwt-allowlist-algorithms" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/oidc/jwt-anatomy/f-jwt-allowlist-algorithms.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;h2 id="evidence"&gt;Evidence &lt;a class="heading-anchor" href="#evidence" aria-label="Link to section: Evidence"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;div class="table-scroll studies" tabindex="0" role="region" aria-label="Specifications"&gt;&lt;table&gt;
&lt;caption&gt;Specifications&lt;/caption&gt;
&lt;thead&gt;&lt;tr&gt;&lt;th scope="col"&gt;Study&lt;/th&gt;&lt;th scope="col"&gt;Year&lt;/th&gt;&lt;th scope="col"&gt;Sample&lt;/th&gt;&lt;th scope="col"&gt;Method&lt;/th&gt;&lt;th scope="col"&gt;Finding&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.rfc-editor.org/rfc/rfc7519" class="ext"&gt;RFC 7519: JSON Web Token (JWT)&lt;/a&gt;&lt;span class="org"&gt;IETF&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2015&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Specification&lt;/td&gt;&lt;td&gt;Defines the JWT format, the registered claim names and the privacy considerations for claims.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.rfc-editor.org/rfc/rfc7515" class="ext"&gt;RFC 7515: JSON Web Signature (JWS)&lt;/a&gt;&lt;span class="org"&gt;IETF&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2015&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Specification&lt;/td&gt;&lt;td&gt;Defines the compact serialization: base64url header, payload and signature joined by dots.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.rfc-editor.org/rfc/rfc8725" class="ext"&gt;RFC 8725: JSON Web Token Best Current Practices&lt;/a&gt;&lt;span class="org"&gt;IETF (BCP 225)&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2020&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Best current practice&lt;/td&gt;&lt;td&gt;Requires algorithm allowlists in libraries and audience, issuer and key validation in applications.&lt;/td&gt;&lt;/tr&gt;
&lt;/tbody&gt;&lt;/table&gt;&lt;/div&gt;
&lt;h2 id="what-to-do"&gt;What to do &lt;a class="heading-anchor" href="#what-to-do" aria-label="Link to section: What to do"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="for-engineers"&gt;For engineers &lt;a class="heading-anchor" href="#for-engineers" aria-label="Link to section: For engineers"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;ul&gt;
&lt;li&gt;Treat a decoded payload as untrusted input until the signature verifies. Pin the algorithms you accept instead of reading &lt;code&gt;alg&lt;/code&gt; from the token, and resolve the key by &lt;code&gt;kid&lt;/code&gt; only from a key set you trust, such as the issuer&amp;rsquo;s &lt;a href="https://www.janbrennenstuhl.eu/jwks/"&gt;JWKS&lt;/a&gt;.&lt;/li&gt;
&lt;li&gt;After the signature, check &lt;code&gt;exp&lt;/code&gt;, &lt;code&gt;nbf&lt;/code&gt;, &lt;code&gt;iss&lt;/code&gt; and &lt;code&gt;aud&lt;/code&gt;. RFC 8725 makes audience validation a must whenever one issuer serves several relying parties.&lt;/li&gt;
&lt;li&gt;Keep secrets and personal data out of signed tokens. If a token has to carry them, use &lt;a href="https://www.janbrennenstuhl.eu/jwe/"&gt;JWE&lt;/a&gt; or keep the data on the server and put a reference in the token.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="for-product-and-ux"&gt;For product and UX &lt;a class="heading-anchor" href="#for-product-and-ux" aria-label="Link to section: For product and UX"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;ul&gt;
&lt;li&gt;Tokens end up in logs, browser storage and support tickets. Anything in the payload should be safe to show to whoever might see it there.&lt;/li&gt;
&lt;li&gt;Short token lifetimes are a product decision as much as a security one: they decide how often a session has to be refreshed.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="try-it"&gt;Try it &lt;a class="heading-anchor" href="#try-it" aria-label="Link to section: Try it"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;section class="explainer" data-explainer="jwt" aria-labelledby="ex-jwt"&gt;
&lt;h3 id="ex-jwt"&gt;Decode a token in your browser&lt;/h3&gt;
&lt;div class="explainer-static"&gt;&lt;p&gt;The example token below has 3 parts. Paste your own token into the decoder to split and decode it. Decoding happens only in your browser; nothing is sent anywhere, and the signature is not verified.&lt;/p&gt;
&lt;p&gt;&lt;code&gt;eyJhbGciOiJFUzI1NiIsInR5cCI6IkpXVCIsImtpZCI6IjIwMjYtMTAifQ.eyJpc3MiOiJodHRwczovL2lkLmV4YW1wbGUuY29tIiwic3ViIjoidXNlci0xMjMiLCJhdWQiOiJodHRwczovL2FwaS5leGFtcGxlLmNvbSIsImV4cCI6MTc5MTkwMDAwMCwiaWF0IjoxNzkxODk2NDAwLCJqdGkiOiI3ZjNjOWEifQ.cGxhY2Vob2xkZXIgc2lnbmF0dXJlLCBub3QgdmFsaWQ&lt;/code&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Header&lt;/strong&gt; decodes to &lt;code&gt;{&amp;quot;alg&amp;quot;:&amp;quot;ES256&amp;quot;,&amp;quot;typ&amp;quot;:&amp;quot;JWT&amp;quot;,&amp;quot;kid&amp;quot;:&amp;quot;2026-10&amp;quot;}&lt;/code&gt;: an ECDSA P-256 signature, and the ID of the key that made it.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Payload&lt;/strong&gt; decodes to &lt;code&gt;{&amp;quot;iss&amp;quot;:&amp;quot;https://id.example.com&amp;quot;,&amp;quot;sub&amp;quot;:&amp;quot;user-123&amp;quot;,&amp;quot;aud&amp;quot;:&amp;quot;https://api.example.com&amp;quot;,&amp;quot;exp&amp;quot;:1791900000,&amp;quot;iat&amp;quot;:1791896400,&amp;quot;jti&amp;quot;:&amp;quot;7f3c9a&amp;quot;}&lt;/code&gt;: issued at 13:00 UTC on 13 October 2026, expiring 1 hour later.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Signature&lt;/strong&gt; is a placeholder in this example. In a real token it is the ES256 signature over the encoded header and payload, checked with the issuer&amp;rsquo;s public key from its JWKS.&lt;/li&gt;
&lt;/ul&gt;
&lt;/div&gt;
&lt;div class="explainer-app" hidden&gt;&lt;/div&gt;
&lt;/section&gt;
&lt;h2 id="how-the-parts-fit-together"&gt;How the parts fit together &lt;a class="heading-anchor" href="#how-the-parts-fit-together" aria-label="Link to section: How the parts fit together"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;p&gt;The header and payload are JSON objects. Each is encoded with base64url, a URL-safe variant of base64 without padding, and the two results are joined with a dot. The signer signs that string with the algorithm named in the header and appends the base64url-encoded signature after a second dot. That is the JWS compact serialization from RFC 7515.&lt;/p&gt;
&lt;p&gt;A verifier reverses the steps: split at the dots, decode the header, pick the key and algorithm it allows, and verify the signature over the first 2 parts exactly as received. Only then does it decode the payload and evaluate the claims. The essays on &lt;a href="https://www.janbrennenstuhl.eu/jwt-stateless-authentication/"&gt;the purpose of JWT&lt;/a&gt;, &lt;a href="https://www.janbrennenstuhl.eu/jwa/"&gt;JSON Web Algorithms&lt;/a&gt; and &lt;a href="https://www.janbrennenstuhl.eu/jwt-vs-oauth/"&gt;JWT versus OAuth 2.0&lt;/a&gt; go deeper.&lt;/p&gt;
&lt;h2 id="related"&gt;Related &lt;a class="heading-anchor" href="#related" aria-label="Link to section: Related"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/token-exchange/"&gt;How does OAuth 2.0 Token Exchange keep provider tokens away from agents?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/oidc/openid-connect-security/"&gt;Where do OpenID Connect implementations break?&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>How does OAuth 2.0 Token Exchange keep provider tokens away from agents?</title><link>https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/token-exchange/</link><guid isPermaLink="true">https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/token-exchange/</guid><pubDate>Tue, 13 Oct 2026 00:00:00 +0200</pubDate><description>&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; OAuth 2.0 Token Exchange (RFC 8693) lets a client trade one token for another token meant for a specific target service. When a gateway runs that exchange on each tool call, as in the Agentic Identity Broker, the provider token stays between the broker, the gateway and the provider and never reaches the agent.&lt;/p&gt;&lt;h2 id="key-findings"&gt;Key findings &lt;a class="heading-anchor" href="#key-findings" aria-label="Link to section: Key findings"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;figure class="finding" id="f-agent-rfc8693-security-token-service"&gt;
&lt;blockquote&gt;&lt;p&gt;RFC 8693 lets a client present a security token to an OAuth 2.0 authorization server acting as a Security Token Service and receive a new token, for example one scoped more narrowly for a downstream service.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.rfc-editor.org/rfc/rfc8693#section-1" class="ext"&gt;&lt;cite&gt;RFC 8693, OAuth 2.0 Token Exchange (IETF, 2020), Section 1&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-agent-rfc8693-security-token-service" title="#f-agent-rfc8693-security-token-service"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="RFC 8693 lets a client present a security token to an OAuth 2.0 authorization server acting as a Security Token Service and receive a new token, for example one scoped more narrowly for a downstream service." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/token-exchange/#f-agent-rfc8693-security-token-service" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/token-exchange/f-agent-rfc8693-security-token-service.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-agent-broker-provider-token-stays-out"&gt;
&lt;blockquote&gt;&lt;p&gt;In the Agentic Identity Broker, an MCP gateway exchanges the agent&amp;rsquo;s token for the matching provider token through RFC 8693 on each tool call, so the provider token never reaches the agent.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.janbrennenstuhl.eu/agentic-identity-broker/" class="ext"&gt;&lt;cite&gt;What I Learned Building an Identity Broker With Coding Agents (janbrennenstuhl.eu, 2026)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-agent-broker-provider-token-stays-out" title="#f-agent-broker-provider-token-stays-out"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="In the Agentic Identity Broker, an MCP gateway exchanges the agent&amp;rsquo;s token for the matching provider token through RFC 8693 on each tool call, so the provider token never reaches the agent." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/token-exchange/#f-agent-broker-provider-token-stays-out" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/token-exchange/f-agent-broker-provider-token-stays-out.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-agent-exchange-client-authentication"&gt;
&lt;blockquote&gt;&lt;p&gt;RFC 8693 warns that without client authentication at the token endpoint, anyone holding a compromised token can trade it for other tokens.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.rfc-editor.org/rfc/rfc8693#section-2.1" class="ext"&gt;&lt;cite&gt;RFC 8693, OAuth 2.0 Token Exchange (IETF, 2020), Section 2.1&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-agent-exchange-client-authentication" title="#f-agent-exchange-client-authentication"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="RFC 8693 warns that without client authentication at the token endpoint, anyone holding a compromised token can trade it for other tokens." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/token-exchange/#f-agent-exchange-client-authentication" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/token-exchange/f-agent-exchange-client-authentication.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-agent-act-claim-current-actor"&gt;
&lt;blockquote&gt;&lt;p&gt;In a JWT issued through token exchange, access control decisions must consider only the top-level claims and the current actor named in the act claim, while nested prior actors are informational.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.rfc-editor.org/rfc/rfc8693#section-4.1" class="ext"&gt;&lt;cite&gt;RFC 8693, OAuth 2.0 Token Exchange (IETF, 2020), Section 4.1&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-agent-act-claim-current-actor" title="#f-agent-act-claim-current-actor"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="In a JWT issued through token exchange, access control decisions must consider only the top-level claims and the current actor named in the act claim, while nested prior actors are informational." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/token-exchange/#f-agent-act-claim-current-actor" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/token-exchange/f-agent-act-claim-current-actor.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-agent-exchange-no-revocation-link"&gt;
&lt;blockquote&gt;&lt;p&gt;A token exchange creates no tight link between the input and output tokens, so passing revocation from one to the other depends on the implementation and isn&amp;rsquo;t a property of the protocol.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.rfc-editor.org/rfc/rfc8693#section-2.1" class="ext"&gt;&lt;cite&gt;RFC 8693, OAuth 2.0 Token Exchange (IETF, 2020), Section 2.1&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-agent-exchange-no-revocation-link" title="#f-agent-exchange-no-revocation-link"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="A token exchange creates no tight link between the input and output tokens, so passing revocation from one to the other depends on the implementation and isn&amp;rsquo;t a property of the protocol." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/token-exchange/#f-agent-exchange-no-revocation-link" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/token-exchange/f-agent-exchange-no-revocation-link.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;h2 id="evidence"&gt;Evidence &lt;a class="heading-anchor" href="#evidence" aria-label="Link to section: Evidence"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;div class="table-scroll studies" tabindex="0" role="region" aria-label="Primary sources on token exchange for agents"&gt;&lt;table&gt;
&lt;caption&gt;Primary sources on token exchange for agents&lt;/caption&gt;
&lt;thead&gt;&lt;tr&gt;&lt;th scope="col"&gt;Study&lt;/th&gt;&lt;th scope="col"&gt;Year&lt;/th&gt;&lt;th scope="col"&gt;Sample&lt;/th&gt;&lt;th scope="col"&gt;Method&lt;/th&gt;&lt;th scope="col"&gt;Finding&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.rfc-editor.org/rfc/rfc8693" class="ext"&gt;OAuth 2.0 Token Exchange (RFC 8693)&lt;/a&gt;&lt;span class="org"&gt;IETF&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2020&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Specification&lt;/td&gt;&lt;td&gt;Defines the token exchange grant, its request and response parameters, and the act and may_act claims.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.rfc-editor.org/rfc/rfc8707" class="ext"&gt;Resource Indicators for OAuth 2.0 (RFC 8707)&lt;/a&gt;&lt;span class="org"&gt;IETF&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2020&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Specification&lt;/td&gt;&lt;td&gt;Defines the resource parameter that tells the authorization server where a token will be used.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.janbrennenstuhl.eu/agentic-identity-broker/" class="ext"&gt;What I Learned Building an Identity Broker With Coding Agents&lt;/a&gt;&lt;span class="org"&gt;this site&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2026&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Field report&lt;/td&gt;&lt;td&gt;Describes how a gateway exchanges the agent&amp;rsquo;s token for the matching provider token on each tool call.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.janbrennenstuhl.eu/talks/economies-of-scale-for-mcp/" class="ext"&gt;Economies of Scale for MCP and Agents&lt;/a&gt;&lt;span class="org"&gt;this site&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2026&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Talk&lt;/td&gt;&lt;td&gt;Presents the broker&amp;rsquo;s delegation model at AGNTCon + MCPCon Europe 2026.&lt;/td&gt;&lt;/tr&gt;
&lt;/tbody&gt;&lt;/table&gt;&lt;/div&gt;
&lt;h2 id="what-to-do"&gt;What to do &lt;a class="heading-anchor" href="#what-to-do" aria-label="Link to section: What to do"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="for-engineers"&gt;For engineers &lt;a class="heading-anchor" href="#for-engineers" aria-label="Link to section: For engineers"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;Put the exchange in a component the agent can&amp;rsquo;t control, such as a gateway, and authenticate that component at the token endpoint. Name the target with &lt;code&gt;resource&lt;/code&gt; or &lt;code&gt;audience&lt;/code&gt; and ask for the smallest &lt;code&gt;scope&lt;/code&gt; the tool call needs. RFC 8693 points out that broad requests across many target services are more likely to be refused.&lt;/p&gt;
&lt;p&gt;Treat revocation as your own job. The RFC leaves it to the implementation whether revoking a grant or an input token affects tokens already issued, so keep issued tokens short-lived and check the grant at exchange time. &lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/consent-and-revocation/"&gt;Consent and revocation&lt;/a&gt; covers the rest.&lt;/p&gt;
&lt;p&gt;If an agent&amp;rsquo;s token can travel further, use the &lt;code&gt;act&lt;/code&gt; claim to record the agent as the actor and the &lt;code&gt;may_act&lt;/code&gt; claim to state who may act for the subject.&lt;/p&gt;
&lt;h3 id="for-product-and-ux"&gt;For product and UX &lt;a class="heading-anchor" href="#for-product-and-ux" aria-label="Link to section: For product and UX"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;Users don&amp;rsquo;t see the exchange, but they see its effects. When a tool call fails because a grant expired or was revoked, tell the user which service and which agent were affected and how to grant access again.&lt;/p&gt;
&lt;h2 id="how-it-works"&gt;How it works &lt;a class="heading-anchor" href="#how-it-works" aria-label="Link to section: How it works"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;p&gt;RFC 8693 defines the exchange as 1 request and 1 response at an authorization server&amp;rsquo;s token endpoint. With a gateway between an agent and its tools, the exchange runs like this:&lt;/p&gt;
&lt;section class="explainer" data-explainer="delegation" aria-labelledby="ex-delegation"&gt;
&lt;h3 id="ex-delegation"&gt;Step through the exchange&lt;/h3&gt;
&lt;div class="explainer-static"&gt;&lt;ol&gt;
&lt;li&gt;The agent calls a tool through the gateway and presents its own token.&lt;/li&gt;
&lt;li&gt;The gateway acts as the RFC 8693 client. It sends a &lt;code&gt;POST&lt;/code&gt; to the token endpoint with &lt;code&gt;grant_type=urn:ietf:params:oauth:grant-type:token-exchange&lt;/code&gt;, a &lt;code&gt;subject_token&lt;/code&gt; that represents the person on whose behalf the request is made, and its &lt;code&gt;subject_token_type&lt;/code&gt;. Optional parameters name the target (&lt;code&gt;resource&lt;/code&gt;, &lt;code&gt;audience&lt;/code&gt;), the access wanted (&lt;code&gt;scope&lt;/code&gt;), and the token type wanted (&lt;code&gt;requested_token_type&lt;/code&gt;).&lt;/li&gt;
&lt;li&gt;The gateway authenticates with the normal OAuth 2.0 client authentication methods.&lt;/li&gt;
&lt;li&gt;The authorization server validates the presented tokens and decides by policy whether to issue a token for the named target. If it refuses, it answers with &lt;code&gt;invalid_request&lt;/code&gt;, or with &lt;code&gt;invalid_target&lt;/code&gt; when it won&amp;rsquo;t issue a token for the requested target.&lt;/li&gt;
&lt;li&gt;A successful response is a normal OAuth token response: &lt;code&gt;access_token&lt;/code&gt; carries the issued token, together with &lt;code&gt;issued_token_type&lt;/code&gt;, &lt;code&gt;token_type&lt;/code&gt;, and, as recommended, &lt;code&gt;expires_in&lt;/code&gt;. The issued token doesn&amp;rsquo;t have to be an OAuth access token.&lt;/li&gt;
&lt;li&gt;The gateway calls the target service with the issued token and returns only the result to the agent.&lt;/li&gt;
&lt;/ol&gt;
&lt;/div&gt;
&lt;div class="explainer-app" hidden&gt;&lt;/div&gt;
&lt;/section&gt;
&lt;p&gt;The following request shows step 2 with placeholder hosts and tokens. It illustrates the RFC parameters and isn&amp;rsquo;t the broker&amp;rsquo;s API. I&amp;rsquo;ve left the values unencoded and added line breaks for readability; a real request URL-encodes the form body.&lt;/p&gt;
&lt;figure class="code" data-lang="http"&gt;&lt;figcaption class="code-file"&gt;&lt;span class="sr-only"&gt;File: &lt;/span&gt;token-exchange.http&lt;/figcaption&gt;&lt;div class="highlight"&gt;&lt;pre tabindex="0" aria-label="token-exchange.http" class="chroma"&gt;&lt;code data-lang="http"&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="nf"&gt;POST&lt;/span&gt; &lt;span class="nn"&gt;/token&lt;/span&gt; &lt;span class="kr"&gt;HTTP&lt;/span&gt;&lt;span class="o"&gt;/&lt;/span&gt;&lt;span class="m"&gt;1.1&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="n"&gt;Host&lt;/span&gt;&lt;span class="o"&gt;:&lt;/span&gt; &lt;span class="l"&gt;broker.example.com&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="n"&gt;Authorization&lt;/span&gt;&lt;span class="o"&gt;:&lt;/span&gt; &lt;span class="l"&gt;Basic &amp;lt;gateway client credentials&amp;gt;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="n"&gt;Content-Type&lt;/span&gt;&lt;span class="o"&gt;:&lt;/span&gt; &lt;span class="l"&gt;application/x-www-form-urlencoded&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;grant_type=urn:ietf:params:oauth:grant-type:token-exchange
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&amp;amp;subject_token=&amp;lt;token the agent presented&amp;gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&amp;amp;subject_token_type=urn:ietf:params:oauth:token-type:access_token
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&amp;amp;resource=https://api.provider.example/
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&amp;amp;scope=&amp;lt;permissions the tool call needs&amp;gt;&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;
&lt;/figure&gt;
&lt;p&gt;The Agentic Identity Broker that Magnus Jungsbluth and I built uses this pattern, as I describe in the &lt;a href="https://www.janbrennenstuhl.eu/agentic-identity-broker/"&gt;retrospective&lt;/a&gt;. A user grants an agent access to a service, such as GitHub, with a chosen permission set and an optional expiry, and the broker stores the provider tokens encrypted. On each tool call, an MCP gateway such as agentgateway exchanges the agent&amp;rsquo;s token for the matching provider token, so the provider token never reaches the agent.&lt;/p&gt;
&lt;p&gt;RFC 8693 also defines an optional &lt;code&gt;actor_token&lt;/code&gt; for the party that will act on behalf of the subject. In the issued JWT, the &lt;code&gt;act&lt;/code&gt; claim can record that actor, and nested &lt;code&gt;act&lt;/code&gt; claims record a chain of earlier actors. The &lt;a href="https://github.com/zalando-incubator/agentic-identity-broker#roadmap" title="Agentic Identity Broker roadmap" class="ext"&gt;broker roadmap&lt;/a&gt; lists token exchange for agent-to-agent calls as planned work.&lt;/p&gt;
&lt;h2 id="related"&gt;Related &lt;a class="heading-anchor" href="#related" aria-label="Link to section: Related"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/delegation-vs-robot-accounts/"&gt;Should an agent use a robot account or delegated access?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/mcp-authorization/"&gt;How does authorization work for MCP servers?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/talks/economies-of-scale-for-mcp/"&gt;Economies of Scale for MCP and Agents&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/projects/agentic-identity-broker/"&gt;Agentic Identity Broker&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>How should people grant and revoke an agent's access?</title><link>https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/consent-and-revocation/</link><guid isPermaLink="true">https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/consent-and-revocation/</guid><pubDate>Tue, 13 Oct 2026 00:00:00 +0200</pubDate><description>&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; Let people grant each agent access per service, with a permission set they choose and an expiry, and let them revoke that grant at any time. Revoking the grant has to stop every token derived from it, so keep issued tokens short-lived and check the grant whenever a new token is issued.&lt;/p&gt;&lt;h2 id="key-findings"&gt;Key findings &lt;a class="heading-anchor" href="#key-findings" aria-label="Link to section: Key findings"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;figure class="finding" id="f-agent-grant-scope-expiry-revoke"&gt;
&lt;blockquote&gt;&lt;p&gt;In the Agentic Identity Broker, a user grants an agent access to a service with a chosen permission set and an optional expiry, and can revoke that grant at any time.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.janbrennenstuhl.eu/agentic-identity-broker/" class="ext"&gt;&lt;cite&gt;What I Learned Building an Identity Broker With Coding Agents (janbrennenstuhl.eu, 2026)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-agent-grant-scope-expiry-revoke" title="#f-agent-grant-scope-expiry-revoke"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="In the Agentic Identity Broker, a user grants an agent access to a service with a chosen permission set and an optional expiry, and can revoke that grant at any time." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/consent-and-revocation/#f-agent-grant-scope-expiry-revoke" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/consent-and-revocation/f-agent-grant-scope-expiry-revoke.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-agent-revocation-cascades-to-grant"&gt;
&lt;blockquote&gt;&lt;p&gt;Under RFC 7009, revoking a refresh token should also invalidate all access tokens based on the same authorization grant if the authorization server supports access token revocation.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.rfc-editor.org/rfc/rfc7009#section-2.1" class="ext"&gt;&lt;cite&gt;RFC 7009, OAuth 2.0 Token Revocation (IETF, 2013), Section 2.1&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-agent-revocation-cascades-to-grant" title="#f-agent-revocation-cascades-to-grant"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Under RFC 7009, revoking a refresh token should also invalidate all access tokens based on the same authorization grant if the authorization server supports access token revocation." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/consent-and-revocation/#f-agent-revocation-cascades-to-grant" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/consent-and-revocation/f-agent-revocation-cascades-to-grant.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-agent-self-contained-token-revocation"&gt;
&lt;blockquote&gt;&lt;p&gt;Immediate revocation of self-contained access tokens needs backend coordination that RFC 7009 calls non-standardized, and short-lived access tokens limit how long a revoked token stays in use.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.rfc-editor.org/rfc/rfc7009#section-3" class="ext"&gt;&lt;cite&gt;RFC 7009, OAuth 2.0 Token Revocation (IETF, 2013), Section 3&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-agent-self-contained-token-revocation" title="#f-agent-self-contained-token-revocation"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Immediate revocation of self-contained access tokens needs backend coordination that RFC 7009 calls non-standardized, and short-lived access tokens limit how long a revoked token stays in use." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/consent-and-revocation/#f-agent-self-contained-token-revocation" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/consent-and-revocation/f-agent-self-contained-token-revocation.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-agent-refresh-token-bound-to-consent"&gt;
&lt;blockquote&gt;&lt;p&gt;Refresh tokens must be bound to the scope and resource servers the resource owner consented to.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.rfc-editor.org/rfc/rfc9700#section-4.14.2" class="ext"&gt;&lt;cite&gt;RFC 9700, Best Current Practice for OAuth 2.0 Security (IETF, 2025), Section 4.14.2&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-agent-refresh-token-bound-to-consent" title="#f-agent-refresh-token-bound-to-consent"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Refresh tokens must be bound to the scope and resource servers the resource owner consented to." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/consent-and-revocation/#f-agent-refresh-token-bound-to-consent" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/consent-and-revocation/f-agent-refresh-token-bound-to-consent.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;h2 id="evidence"&gt;Evidence &lt;a class="heading-anchor" href="#evidence" aria-label="Link to section: Evidence"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;div class="table-scroll studies" tabindex="0" role="region" aria-label="Primary sources on agent consent and revocation"&gt;&lt;table&gt;
&lt;caption&gt;Primary sources on agent consent and revocation&lt;/caption&gt;
&lt;thead&gt;&lt;tr&gt;&lt;th scope="col"&gt;Study&lt;/th&gt;&lt;th scope="col"&gt;Year&lt;/th&gt;&lt;th scope="col"&gt;Sample&lt;/th&gt;&lt;th scope="col"&gt;Method&lt;/th&gt;&lt;th scope="col"&gt;Finding&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.rfc-editor.org/rfc/rfc7009" class="ext"&gt;OAuth 2.0 Token Revocation (RFC 7009)&lt;/a&gt;&lt;span class="org"&gt;IETF&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2013&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Specification&lt;/td&gt;&lt;td&gt;Defines a revocation endpoint and how revoking one token can revoke related tokens and the grant.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.rfc-editor.org/rfc/rfc9700" class="ext"&gt;Best Current Practice for OAuth 2.0 Security (RFC 9700)&lt;/a&gt;&lt;span class="org"&gt;IETF&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2025&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Specification&lt;/td&gt;&lt;td&gt;Binds refresh tokens to the consented scope and resources and lets them expire after inactivity.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.rfc-editor.org/rfc/rfc8707" class="ext"&gt;Resource Indicators for OAuth 2.0 (RFC 8707)&lt;/a&gt;&lt;span class="org"&gt;IETF&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2020&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Specification&lt;/td&gt;&lt;td&gt;Lets the authorization server tell the user which resources a client will access on their behalf.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.rfc-editor.org/rfc/rfc8693" class="ext"&gt;OAuth 2.0 Token Exchange (RFC 8693)&lt;/a&gt;&lt;span class="org"&gt;IETF&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2020&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Specification&lt;/td&gt;&lt;td&gt;Leaves revocation of tokens issued by an exchange to the implementation.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://modelcontextprotocol.io/specification/2026-07-28/basic/authorization" class="ext"&gt;Model Context Protocol: Authorization (revision 2026-07-28)&lt;/a&gt;&lt;span class="org"&gt;Model Context Protocol&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2026&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Specification&lt;/td&gt;&lt;td&gt;Asks clients to request minimal scopes first and add scopes through step-up authorization.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.janbrennenstuhl.eu/agentic-identity-broker/" class="ext"&gt;What I Learned Building an Identity Broker With Coding Agents&lt;/a&gt;&lt;span class="org"&gt;this site&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2026&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Field report&lt;/td&gt;&lt;td&gt;Describes per-agent grants with chosen permissions, optional expiry and revocation at any time.&lt;/td&gt;&lt;/tr&gt;
&lt;/tbody&gt;&lt;/table&gt;&lt;/div&gt;
&lt;h2 id="what-to-do"&gt;What to do &lt;a class="heading-anchor" href="#what-to-do" aria-label="Link to section: What to do"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="for-engineers"&gt;For engineers &lt;a class="heading-anchor" href="#for-engineers" aria-label="Link to section: For engineers"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;Model the grant as a record of its own that names the person, the agent, the service, the permissions and the expiry, and issue every token from it. Check that record whenever you issue or exchange a token, so a revocation takes effect at the next tool call. Expiring tokens are the fallback for tokens you can&amp;rsquo;t recall.&lt;/p&gt;
&lt;p&gt;Wire revocation through every layer you control. Call the &lt;a href="https://www.rfc-editor.org/rfc/rfc7009" class="ext"&gt;RFC 7009&lt;/a&gt; revocation endpoint where the provider offers one, and don&amp;rsquo;t stop at the first token. RFC 8693 says tokens issued by a &lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/token-exchange/"&gt;token exchange&lt;/a&gt; aren&amp;rsquo;t linked to their input token, so revoking one doesn&amp;rsquo;t reach the other unless you build that.&lt;/p&gt;
&lt;p&gt;Ask for the least access first. The &lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/mcp-authorization/"&gt;MCP authorization specification&lt;/a&gt; describes this as step-up authorization: request minimal scopes and add more when a server asks for them.&lt;/p&gt;
&lt;h3 id="for-product-and-ux"&gt;For product and UX &lt;a class="heading-anchor" href="#for-product-and-ux" aria-label="Link to section: For product and UX"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;Show each grant with the agent, the service, the permissions and the expiry in plain words, and put the revoke action next to it. Default to an expiry instead of open-ended access. When an agent needs more access, ask at that moment for that specific permission.&lt;/p&gt;
&lt;h2 id="how-it-works"&gt;How it works &lt;a class="heading-anchor" href="#how-it-works" aria-label="Link to section: How it works"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;p&gt;Granting access in OAuth starts with an authorization request. The client names the access it wants in &lt;code&gt;scope&lt;/code&gt; and, with &lt;a href="https://www.rfc-editor.org/rfc/rfc8707#section-2.1" class="ext"&gt;RFC 8707&lt;/a&gt;, the resources in &lt;code&gt;resource&lt;/code&gt;. RFC 8707 notes that the authorization server can use those resource values to inform the user about the resources the client will access on their behalf. The consent screen can then name the service, not only a list of scopes.&lt;/p&gt;
&lt;p&gt;In the Agentic Identity Broker, which Magnus Jungsbluth and I built, a user grants an agent access to a service with a chosen permission set and an optional expiry. The broker keeps both the user and the agent in every authorization decision. I describe the design in the &lt;a href="https://www.janbrennenstuhl.eu/agentic-identity-broker/"&gt;retrospective&lt;/a&gt;, and my &lt;a href="https://www.janbrennenstuhl.eu/appsec-agentic-engineering/"&gt;AppSec essay&lt;/a&gt; argues for narrow, time-limited permissions that the user can revoke at any moment.&lt;/p&gt;
&lt;p&gt;Revoking access in OAuth uses the revocation endpoint from RFC 7009. A client sends the token it wants revoked, and the server invalidates it and, depending on policy, related tokens and the grant itself. RFC 7009 gives 2 reasons this matters for people: it stops a grant from staying valid without the user&amp;rsquo;s knowledge, and revoked grants disappear from any list of grants the authorization server shows the user. The server answers with HTTP 200 even for an invalid token. A 503 means the client must assume the token still exists.&lt;/p&gt;
&lt;p&gt;Self-contained tokens make revocation harder. A resource server can check a self-contained token, such as a JSON Web Token (JWT), without asking the authorization server, so it won&amp;rsquo;t learn about a revocation on its own. RFC 7009 names short-lived access tokens with refresh tokens as one alternative. RFC 9700 adds that refresh tokens must stay bound to the consented scope and resources, should expire after a period of inactivity, and may be revoked on security events such as a password change or logout.&lt;/p&gt;
&lt;h2 id="related"&gt;Related &lt;a class="heading-anchor" href="#related" aria-label="Link to section: Related"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/delegation-vs-robot-accounts/"&gt;Should an agent use a robot account or delegated access?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/token-exchange/"&gt;How does OAuth 2.0 Token Exchange keep provider tokens away from agents?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/talks/economies-of-scale-for-mcp/"&gt;Economies of Scale for MCP and Agents&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/projects/agentic-identity-broker/"&gt;Agentic Identity Broker&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>How does authorization work for MCP servers?</title><link>https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/mcp-authorization/</link><guid isPermaLink="true">https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/mcp-authorization/</guid><pubDate>Tue, 13 Oct 2026 00:00:00 +0200</pubDate><description>&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; In the MCP authorization specification (revision 2026-07-28), a protected MCP server over HTTP is an OAuth 2.1 resource server: it publishes Protected Resource Metadata that points clients to its authorization server and accepts only tokens issued for itself. MCP clients discover that authorization server, run the authorization code flow with PKCE, and name the MCP server in a resource parameter on every authorization and token request.&lt;/p&gt;&lt;p&gt;This entry describes the Model Context Protocol (MCP) &lt;a href="https://modelcontextprotocol.io/specification/2026-07-28/basic/authorization" class="ext"&gt;authorization specification&lt;/a&gt; as of revision 2026-07-28, which the MCP versioning page listed as the current revision on 13 October 2026. Later revisions may change these requirements.&lt;/p&gt;
&lt;h2 id="key-findings"&gt;Key findings &lt;a class="heading-anchor" href="#key-findings" aria-label="Link to section: Key findings"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;figure class="finding" id="f-agent-mcp-protected-resource-metadata"&gt;
&lt;blockquote&gt;&lt;p&gt;As of MCP revision 2026-07-28, MCP servers must implement OAuth 2.0 Protected Resource Metadata (RFC 9728), and MCP clients must use it to discover the authorization server.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://modelcontextprotocol.io/specification/2026-07-28/basic/authorization" class="ext"&gt;&lt;cite&gt;Model Context Protocol, Authorization, revision 2026-07-28&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-agent-mcp-protected-resource-metadata" title="#f-agent-mcp-protected-resource-metadata"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="As of MCP revision 2026-07-28, MCP servers must implement OAuth 2.0 Protected Resource Metadata (RFC 9728), and MCP clients must use it to discover the authorization server." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/mcp-authorization/#f-agent-mcp-protected-resource-metadata" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/mcp-authorization/f-agent-mcp-protected-resource-metadata.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-agent-mcp-resource-parameter"&gt;
&lt;blockquote&gt;&lt;p&gt;MCP clients must send the RFC 8707 resource parameter with the MCP server&amp;rsquo;s canonical URI in both authorization and token requests, whether or not the authorization server supports it.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://modelcontextprotocol.io/specification/2026-07-28/basic/authorization" class="ext"&gt;&lt;cite&gt;Model Context Protocol, Authorization, revision 2026-07-28&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-agent-mcp-resource-parameter" title="#f-agent-mcp-resource-parameter"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="MCP clients must send the RFC 8707 resource parameter with the MCP server&amp;rsquo;s canonical URI in both authorization and token requests, whether or not the authorization server supports it." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/mcp-authorization/#f-agent-mcp-resource-parameter" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/mcp-authorization/f-agent-mcp-resource-parameter.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-agent-mcp-audience-validation"&gt;
&lt;blockquote&gt;&lt;p&gt;MCP servers must validate that an access token was issued specifically for them as the intended audience and must not accept or transit any other tokens.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://modelcontextprotocol.io/specification/2026-07-28/basic/authorization" class="ext"&gt;&lt;cite&gt;Model Context Protocol, Authorization, revision 2026-07-28&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-agent-mcp-audience-validation" title="#f-agent-mcp-audience-validation"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="MCP servers must validate that an access token was issued specifically for them as the intended audience and must not accept or transit any other tokens." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/mcp-authorization/#f-agent-mcp-audience-validation" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/mcp-authorization/f-agent-mcp-audience-validation.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-agent-mcp-no-token-passthrough"&gt;
&lt;blockquote&gt;&lt;p&gt;An MCP server that calls upstream APIs must use a separate token from the upstream authorization server and must not pass through the token it received from the MCP client.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://modelcontextprotocol.io/specification/2026-07-28/basic/authorization/security-considerations" class="ext"&gt;&lt;cite&gt;Model Context Protocol, Authorization Security Considerations, revision 2026-07-28&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-agent-mcp-no-token-passthrough" title="#f-agent-mcp-no-token-passthrough"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="An MCP server that calls upstream APIs must use a separate token from the upstream authorization server and must not pass through the token it received from the MCP client." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/mcp-authorization/#f-agent-mcp-no-token-passthrough" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/mcp-authorization/f-agent-mcp-no-token-passthrough.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;h2 id="evidence"&gt;Evidence &lt;a class="heading-anchor" href="#evidence" aria-label="Link to section: Evidence"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;div class="table-scroll studies" tabindex="0" role="region" aria-label="Primary sources on MCP authorization"&gt;&lt;table&gt;
&lt;caption&gt;Primary sources on MCP authorization&lt;/caption&gt;
&lt;thead&gt;&lt;tr&gt;&lt;th scope="col"&gt;Study&lt;/th&gt;&lt;th scope="col"&gt;Year&lt;/th&gt;&lt;th scope="col"&gt;Sample&lt;/th&gt;&lt;th scope="col"&gt;Method&lt;/th&gt;&lt;th scope="col"&gt;Finding&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://modelcontextprotocol.io/specification/2026-07-28/basic/authorization" class="ext"&gt;Model Context Protocol: Authorization (revision 2026-07-28)&lt;/a&gt;&lt;span class="org"&gt;Model Context Protocol&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2026&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Specification&lt;/td&gt;&lt;td&gt;Defines roles, discovery, the authorization flow, the resource parameter, token handling and step-up authorization for HTTP transports.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://modelcontextprotocol.io/specification/2026-07-28/basic/authorization/security-considerations" class="ext"&gt;Model Context Protocol: Authorization Security Considerations (revision 2026-07-28)&lt;/a&gt;&lt;span class="org"&gt;Model Context Protocol&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2026&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Specification&lt;/td&gt;&lt;td&gt;Requires audience binding, PKCE, exact redirect URI matching and a ban on token passthrough.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.rfc-editor.org/rfc/rfc9728" class="ext"&gt;OAuth 2.0 Protected Resource Metadata (RFC 9728)&lt;/a&gt;&lt;span class="org"&gt;IETF&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2025&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Specification&lt;/td&gt;&lt;td&gt;Defines the metadata document and the WWW-Authenticate parameter that point a client to a resource&amp;rsquo;s authorization servers.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.rfc-editor.org/rfc/rfc8707" class="ext"&gt;Resource Indicators for OAuth 2.0 (RFC 8707)&lt;/a&gt;&lt;span class="org"&gt;IETF&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2020&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Specification&lt;/td&gt;&lt;td&gt;Defines the resource parameter and asks authorization servers to audience-restrict tokens to it.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.janbrennenstuhl.eu/talks/economies-of-scale-for-mcp/" class="ext"&gt;Economies of Scale for MCP and Agents&lt;/a&gt;&lt;span class="org"&gt;this site&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2026&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Talk&lt;/td&gt;&lt;td&gt;States that an agent should not hold a shared, long-lived provider credential and presents the broker&amp;rsquo;s delegation model.&lt;/td&gt;&lt;/tr&gt;
&lt;/tbody&gt;&lt;/table&gt;&lt;/div&gt;
&lt;h2 id="what-to-do"&gt;What to do &lt;a class="heading-anchor" href="#what-to-do" aria-label="Link to section: What to do"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="for-engineers"&gt;For engineers &lt;a class="heading-anchor" href="#for-engineers" aria-label="Link to section: For engineers"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;For an MCP server over HTTP, publish Protected Resource Metadata, answer unauthenticated requests with a 401 that points to it, and check the audience of every token. Put the scopes an operation needs in the &lt;code&gt;WWW-Authenticate&lt;/code&gt; challenge, all at once. If the server calls other APIs, get its own token for each of them. The &lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/token-exchange/"&gt;token exchange entry&lt;/a&gt; shows one way to do that.&lt;/p&gt;
&lt;p&gt;For an MCP client, implement both discovery mechanisms, PKCE with &lt;code&gt;S256&lt;/code&gt;, the &lt;code&gt;resource&lt;/code&gt; parameter, and the issuer check. Never put tokens in the query string.&lt;/p&gt;
&lt;p&gt;An MCP server on the STDIO transport is outside this specification, which says such servers should take credentials from the environment. Keep those credentials short-lived and task-scoped, as my &lt;a href="https://www.janbrennenstuhl.eu/appsec-agentic-engineering/"&gt;AppSec essay&lt;/a&gt; recommends for coding agents.&lt;/p&gt;
&lt;h3 id="for-product-and-ux"&gt;For product and UX &lt;a class="heading-anchor" href="#for-product-and-ux" aria-label="Link to section: For product and UX"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;Start with the minimal scopes and ask for more only when an operation needs them. The specification warns that misconfigured scopes may require frequent user interaction, and that challenging one scope at a time forces several authorization round-trips for one operation.&lt;/p&gt;
&lt;h2 id="how-it-works"&gt;How it works &lt;a class="heading-anchor" href="#how-it-works" aria-label="Link to section: How it works"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;p&gt;The specification defines 3 roles. The MCP server is an OAuth 2.1 resource server, the MCP client is an OAuth 2.1 client acting for a resource owner, and the authorization server issues the tokens. Authorization is optional for MCP implementations. Implementations on HTTP-based transports should conform to the specification, and STDIO implementations should not.&lt;/p&gt;
&lt;p&gt;A first connection follows these steps:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;The client sends an MCP request without a token. The server answers with HTTP 401 and a &lt;code&gt;WWW-Authenticate&lt;/code&gt; header that carries the &lt;code&gt;resource_metadata&lt;/code&gt; URL from RFC 9728 and, as the specification recommends, the required &lt;code&gt;scope&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;The client fetches the Protected Resource Metadata and picks an authorization server from it. It then reads that server&amp;rsquo;s metadata through OAuth 2.0 Authorization Server Metadata (RFC 8414) or OpenID Connect Discovery. Clients must support both, and if the metadata doesn&amp;rsquo;t list &lt;code&gt;code_challenge_methods_supported&lt;/code&gt;, the client must refuse to proceed.&lt;/li&gt;
&lt;li&gt;The client gets a client ID. It uses pre-registered client information if it has some, then Client ID Metadata Documents if the authorization server supports them. Dynamic Client Registration is deprecated and kept only as a fallback for backwards compatibility.&lt;/li&gt;
&lt;li&gt;The client starts the authorization code flow with PKCE, the &lt;code&gt;resource&lt;/code&gt; parameter and the selected scopes, and records the expected issuer. When the response comes back, it validates the &lt;code&gt;iss&lt;/code&gt; parameter (RFC 9207) before it redeems the code.&lt;/li&gt;
&lt;li&gt;The client redeems the code with the &lt;code&gt;code_verifier&lt;/code&gt; and the &lt;code&gt;resource&lt;/code&gt; parameter, then sends the access token in the &lt;code&gt;Authorization: Bearer&lt;/code&gt; header on every request.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;When a token lacks a scope at runtime, the server answers with 403 and &lt;code&gt;error=&amp;quot;insufficient_scope&amp;quot;&lt;/code&gt;. The client then runs a step-up flow with the union of its previous scopes and the new ones.&lt;/p&gt;
&lt;p&gt;Revision 2026-07-28 doesn&amp;rsquo;t cover token exchange in its core authorization pages. Optional authorization extensions live in a separate repository. In the Agentic Identity Broker, which Magnus Jungsbluth and I built, an MCP gateway exchanges the agent&amp;rsquo;s token for the matching provider token through RFC 8693, so the provider token never reaches the agent. I describe that design in the &lt;a href="https://www.janbrennenstuhl.eu/agentic-identity-broker/"&gt;retrospective&lt;/a&gt;, and we presented it in our &lt;a href="https://www.janbrennenstuhl.eu/talks/economies-of-scale-for-mcp/"&gt;AGNTCon + MCPCon talk&lt;/a&gt;.&lt;/p&gt;
&lt;h2 id="related"&gt;Related &lt;a class="heading-anchor" href="#related" aria-label="Link to section: Related"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/delegation-vs-robot-accounts/"&gt;Should an agent use a robot account or delegated access?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/agent-identity/consent-and-revocation/"&gt;How should people grant and revoke an agent&amp;rsquo;s access?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/projects/agentic-identity-broker/"&gt;Agentic Identity Broker&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>How do attackers take over accounts?</title><link>https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-takeover/</link><guid isPermaLink="true">https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-takeover/</guid><pubDate>Fri, 21 Apr 2023 00:00:00 +0200</pubDate><description>&lt;p&gt;&lt;em&gt;Latest change (2026-10-13): Rewritten as a field-guide entry: question title, short answer, key findings and evidence table. Study summaries kept and extended with numbers from the papers.&lt;/em&gt;&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; Some attackers prepare an account before the victim even signs up: at least 35 of 75 popular services studied were vulnerable to such pre-hijacking. Others are paid to break into a specific account and mostly use tailored phishing that can get past SMS two-factor authentication.&lt;/p&gt;&lt;h2 id="key-findings"&gt;Key findings &lt;a class="heading-anchor" href="#key-findings" aria-label="Link to section: Key findings"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;figure class="finding" id="f-hijack-prehijack-35-of-75"&gt;
&lt;blockquote&gt;&lt;p&gt;At least 35 of 75 popular online services were vulnerable to one or more account pre-hijacking attacks.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://arxiv.org/pdf/2205.10174.pdf" class="ext"&gt;&lt;cite&gt;Sudhodanan and Paverd, Pre-hijacked accounts (USENIX Security 2022)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-hijack-prehijack-35-of-75" title="#f-hijack-prehijack-35-of-75"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="At least 35 of 75 popular online services were vulnerable to one or more account pre-hijacking attacks." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-takeover/#f-hijack-prehijack-35-of-75" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-takeover/f-hijack-prehijack-35-of-75.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-hijack-prehijack-email-only"&gt;
&lt;blockquote&gt;&lt;p&gt;In an account pre-hijacking attack, an attacker who knows only the victim&amp;rsquo;s email address acts before the victim creates an account and gains access after the victim has signed up.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://arxiv.org/pdf/2205.10174.pdf" class="ext"&gt;&lt;cite&gt;Sudhodanan and Paverd, Pre-hijacked accounts (USENIX Security 2022)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-hijack-prehijack-email-only" title="#f-hijack-prehijack-email-only"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="In an account pre-hijacking attack, an attacker who knows only the victim&amp;rsquo;s email address acts before the victim creates an account and gains access after the victim has signed up." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-takeover/#f-hijack-prehijack-email-only" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-takeover/f-hijack-prehijack-email-only.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-hijack-hire-5-of-27"&gt;
&lt;blockquote&gt;&lt;p&gt;Of 27 hack-for-hire services that researchers approached as buyers, only 5 succeeded in attacking the email accounts of the test identities.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://dl.acm.org/doi/pdf/10.1145/3308558.3313489" class="ext"&gt;&lt;cite&gt;Mirian et al., Hack for Hire (WWW 2019)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-hijack-hire-5-of-27" title="#f-hijack-hire-5-of-27"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Of 27 hack-for-hire services that researchers approached as buyers, only 5 succeeded in attacking the email accounts of the test identities." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-takeover/#f-hijack-hire-5-of-27" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-takeover/f-hijack-hire-5-of-27.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-hijack-hire-phishing-sms-2fa"&gt;
&lt;blockquote&gt;&lt;p&gt;Hack-for-hire attackers relied mainly on tailored phishing messages that were sophisticated enough to bypass SMS two-factor authentication.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://dl.acm.org/doi/pdf/10.1145/3308558.3313489" class="ext"&gt;&lt;cite&gt;Mirian et al., Hack for Hire (WWW 2019)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-hijack-hire-phishing-sms-2fa" title="#f-hijack-hire-phishing-sms-2fa"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Hack-for-hire attackers relied mainly on tailored phishing messages that were sophisticated enough to bypass SMS two-factor authentication." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-takeover/#f-hijack-hire-phishing-sms-2fa" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-takeover/f-hijack-hire-phishing-sms-2fa.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;h2 id="evidence"&gt;Evidence &lt;a class="heading-anchor" href="#evidence" aria-label="Link to section: Evidence"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;div class="table-scroll studies" tabindex="0" role="region" aria-label="Account hijacking studies"&gt;&lt;table&gt;
&lt;caption&gt;Account hijacking studies&lt;/caption&gt;
&lt;thead&gt;&lt;tr&gt;&lt;th scope="col"&gt;Study&lt;/th&gt;&lt;th scope="col"&gt;Year&lt;/th&gt;&lt;th scope="col"&gt;Sample&lt;/th&gt;&lt;th scope="col"&gt;Method&lt;/th&gt;&lt;th scope="col"&gt;Finding&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://arxiv.org/pdf/2205.10174.pdf" class="ext"&gt;Pre-hijacked accounts: An Empirical Study of Security Failures in User Account Creation on the Web&lt;/a&gt;&lt;span class="org"&gt;Independent researcher, Microsoft Security Response Center&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2022&lt;/td&gt;&lt;td&gt;75 popular online services&lt;/td&gt;&lt;td&gt;Attack design and testing against live services&lt;/td&gt;&lt;td&gt;At least 35 services were vulnerable to one or more of 5 pre-hijacking attacks.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://dl.acm.org/doi/pdf/10.1145/3308558.3313489" class="ext"&gt;Hack for Hire: Exploring the Emerging Market for Account Hijacking&lt;/a&gt;&lt;span class="org"&gt;UC San Diego, Google&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2019&lt;/td&gt;&lt;td&gt;27 English, Russian and Chinese hack-for-hire services&lt;/td&gt;&lt;td&gt;Posing as buyers against synthetic victim identities&lt;/td&gt;&lt;td&gt;5 services succeeded, mostly through tailored phishing.&lt;/td&gt;&lt;/tr&gt;
&lt;/tbody&gt;&lt;/table&gt;&lt;/div&gt;
&lt;h2 id="what-to-do"&gt;What to do &lt;a class="heading-anchor" href="#what-to-do" aria-label="Link to section: What to do"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="for-engineers"&gt;For engineers &lt;a class="heading-anchor" href="#for-engineers" aria-label="Link to section: For engineers"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;ul&gt;
&lt;li&gt;Don&amp;rsquo;t let anyone use an account before they have proven they control its identifier. Verify the email address before the account becomes usable, and before you merge a password account with a federated login for the same address.&lt;/li&gt;
&lt;li&gt;Only trust email addresses from identity providers that verify them. One of the 5 attacks relies on a non-verifying identity provider.&lt;/li&gt;
&lt;li&gt;End all sessions and expire pending email-change links when the password is reset or the email address changes. Two of the attacks rely on unexpired sessions and email changes.&lt;/li&gt;
&lt;li&gt;Offer phishing-resistant factors. The hack-for-hire services got past SMS codes with phishing. See &lt;a href="https://www.janbrennenstuhl.eu/passwordless-authentication/"&gt;passwordless authentication&lt;/a&gt; and my &lt;a href="https://www.janbrennenstuhl.eu/fido-passkeys/"&gt;critical notes on passkeys&lt;/a&gt;.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="for-product-and-ux"&gt;For product and UX &lt;a class="heading-anchor" href="#for-product-and-ux" aria-label="Link to section: For product and UX"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;ul&gt;
&lt;li&gt;Show users which sign-in methods are linked to their account, and notify them when one is added.&lt;/li&gt;
&lt;li&gt;When someone signs up with an address that already has a pending or unverified account, guide them through verification instead of silently signing them in.&lt;/li&gt;
&lt;li&gt;Recovery flows are another takeover path. The &lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-recovery/"&gt;account recovery entry&lt;/a&gt; lists the common weaknesses.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="the-studies-in-detail"&gt;The studies in detail &lt;a class="heading-anchor" href="#the-studies-in-detail" aria-label="Link to section: The studies in detail"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="account-pre-hijacking-attacks-security-failures-in-user-account-creation"&gt;Account Pre-Hijacking Attacks: Security Failures in User Account Creation &lt;a class="heading-anchor" href="#account-pre-hijacking-attacks-security-failures-in-user-account-creation" aria-label="Link to section: Account Pre-Hijacking Attacks: Security Failures in User Account Creation"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;&amp;ldquo;The ubiquity of user accounts in websites and online services makes account hijacking a serious security concern. Although previous research has studied various techniques through which an attacker can gain access to a victim&amp;rsquo;s account, relatively little attention has been directed towards the process of account creation.&amp;rdquo;&lt;/p&gt;
&lt;p&gt;In this study, Avinash Sudhodanan and Andrew Paverd show that there is a whole class of &lt;strong&gt;account pre-hijacking attacks&lt;/strong&gt;. The key point: &amp;ldquo;The distinctive feature of these attacks is that the attacker performs some action before the victim creates an account, which makes it trivial for the attacker to gain access after the victim has created/recovered the account.&amp;rdquo;&lt;/p&gt;
&lt;p&gt;The authors assume a realistic attacker who knows only the victim&amp;rsquo;s email address. They describe 5 attacks: Classic-Federated Merge, Unexpired Session, Trojan Identifier, Unexpired Email Change and Non-verifying IdP. Of 75 popular services they analyzed, at least 35 were vulnerable to one or more of them. Some attacks were completely undetectable from the victim&amp;rsquo;s point of view. The paper closes with security requirements that prevent these vulnerabilities.&lt;/p&gt;
&lt;p&gt;&lt;a href="https://arxiv.org/pdf/2205.10174.pdf" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="hack-for-hire-exploring-the-emerging-market-for-account-hijacking"&gt;Hack for Hire: Exploring the Emerging Market for Account Hijacking &lt;a class="heading-anchor" href="#hack-for-hire-exploring-the-emerging-market-for-account-hijacking" aria-label="Link to section: Hack for Hire: Exploring the Emerging Market for Account Hijacking"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;In this paper, researchers from UC San Diego and Google &amp;ldquo;study a segment of targeted attackers known as &amp;lsquo;hack for hire&amp;rsquo; services to understand the playbook that attackers use to gain access to victim accounts.&amp;rdquo;&lt;/p&gt;
&lt;p&gt;Posing as buyers, they contacted 27 English, Russian and Chinese black-market services, and only 5 succeeded in attacking the synthetic but realistic identities the researchers controlled. The attackers mostly used tailored phishing that could bypass SMS two-factor authentication. The authors also observed low volume, poor customer service and multiple scammers in this market. The paper is a good read if you want to understand the commercial account hijacking ecosystem and the techniques criminals use.&lt;/p&gt;
&lt;p&gt;&lt;a href="https://dl.acm.org/doi/pdf/10.1145/3308558.3313489" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h2 id="related"&gt;Related &lt;a class="heading-anchor" href="#related" aria-label="Link to section: Related"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-recovery/"&gt;How secure are email-based password resets?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/reinforced-authentication/"&gt;Do login challenges and risk scoring stop account takeover?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/oidc/openid-connect-security/"&gt;Where do OpenID Connect implementations break?&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>Where do OpenID Connect implementations break?</title><link>https://www.janbrennenstuhl.eu/knowledge-base/oidc/openid-connect-security/</link><guid isPermaLink="true">https://www.janbrennenstuhl.eu/knowledge-base/oidc/openid-connect-security/</guid><pubDate>Mon, 15 May 2023 00:00:00 +0200</pubDate><description>&lt;p&gt;&lt;em&gt;Latest change (2026-10-13): Rewritten as a field-guide entry: question title, short answer, key findings and evidence table. Study summaries kept and extended with numbers from the thesis abstract.&lt;/em&gt;&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; Implementations break mostly where tokens and redirects get checked. Documented attacks on relying parties include token substitution, key confusion and replay, and attacks on providers include redirect URI manipulation, PKCE downgrade and sub claim spoofing. Certification doesn&amp;rsquo;t rule these out, as a 2020 lab analysis of OpenID Certified implementations disclosed 12 flaws.&lt;/p&gt;&lt;h2 id="key-findings"&gt;Key findings &lt;a class="heading-anchor" href="#key-findings" aria-label="Link to section: Key findings"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;figure class="finding" id="f-oidc-certified-12-flaws"&gt;
&lt;blockquote&gt;&lt;p&gt;Testing 6 OpenID Certified identity providers and 8 service providers against 11 service provider attacks and 7 identity provider attacks uncovered 12 implementation flaws.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.nds.ruhr-uni-bochum.de/media/ei/arbeiten/2021/05/03/masterthesis.pdf" class="ext"&gt;&lt;cite&gt;Fries, Security Analysis of Real-Life OpenID Connect Implementations (Master&amp;#39;s thesis, Ruhr-University Bochum, 2020)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-oidc-certified-12-flaws" title="#f-oidc-certified-12-flaws"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Testing 6 OpenID Certified identity providers and 8 service providers against 11 service provider attacks and 7 identity provider attacks uncovered 12 implementation flaws." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/oidc/openid-connect-security/#f-oidc-certified-12-flaws" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/oidc/openid-connect-security/f-oidc-certified-12-flaws.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-oidc-spec-flaws"&gt;
&lt;blockquote&gt;&lt;p&gt;Mladenov and Mainka classify session overwriting and IdP confusion as flaws in the OpenID Connect specification itself, separate from implementation bugs.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.nds.ruhr-uni-bochum.de/media/ei/veroeffentlichungen/2017/01/13/OIDCSecurity_1.pdf" class="ext"&gt;&lt;cite&gt;Mladenov and Mainka, OpenID Connect Security Considerations (Ruhr-University Bochum, 2017)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-oidc-spec-flaws" title="#f-oidc-spec-flaws"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Mladenov and Mainka classify session overwriting and IdP confusion as flaws in the OpenID Connect specification itself, separate from implementation bugs." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/oidc/openid-connect-security/#f-oidc-spec-flaws" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/oidc/openid-connect-security/f-oidc-spec-flaws.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-oidc-attacker-needs-no-network"&gt;
&lt;blockquote&gt;&lt;p&gt;The attacks in the OpenID Connect Security Considerations report work for a web attacker who sends ordinary HTTP requests and has no control over the network.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.nds.ruhr-uni-bochum.de/media/ei/veroeffentlichungen/2017/01/13/OIDCSecurity_1.pdf" class="ext"&gt;&lt;cite&gt;Mladenov and Mainka, OpenID Connect Security Considerations (Ruhr-University Bochum, 2017)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-oidc-attacker-needs-no-network" title="#f-oidc-attacker-needs-no-network"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="The attacks in the OpenID Connect Security Considerations report work for a web attacker who sends ordinary HTTP requests and has no control over the network." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/oidc/openid-connect-security/#f-oidc-attacker-needs-no-network" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/oidc/openid-connect-security/f-oidc-attacker-needs-no-network.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;h2 id="evidence"&gt;Evidence &lt;a class="heading-anchor" href="#evidence" aria-label="Link to section: Evidence"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;div class="table-scroll studies" tabindex="0" role="region" aria-label="OpenID Connect security studies"&gt;&lt;table&gt;
&lt;caption&gt;OpenID Connect security studies&lt;/caption&gt;
&lt;thead&gt;&lt;tr&gt;&lt;th scope="col"&gt;Study&lt;/th&gt;&lt;th scope="col"&gt;Year&lt;/th&gt;&lt;th scope="col"&gt;Sample&lt;/th&gt;&lt;th scope="col"&gt;Method&lt;/th&gt;&lt;th scope="col"&gt;Finding&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.nds.ruhr-uni-bochum.de/media/ei/arbeiten/2021/05/03/masterthesis.pdf" class="ext"&gt;Security Analysis of Real-Life OpenID Connect Implementations&lt;/a&gt;&lt;span class="org"&gt;Ruhr-University Bochum&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2020&lt;/td&gt;&lt;td&gt;6 identity providers, 8 service providers&lt;/td&gt;&lt;td&gt;Docker lab with PrOfESSOS, automatic and manual tests&lt;/td&gt;&lt;td&gt;18 attack types tested; 12 implementation flaws disclosed.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.nds.ruhr-uni-bochum.de/media/ei/veroeffentlichungen/2017/01/13/OIDCSecurity_1.pdf" class="ext"&gt;OpenID Connect Security Considerations&lt;/a&gt;&lt;span class="org"&gt;Ruhr-University Bochum&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2017&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Technical report on specification and implementation flaws&lt;/td&gt;&lt;td&gt;Catalogues attacks on clients and providers with defence recommendations.&lt;/td&gt;&lt;/tr&gt;
&lt;/tbody&gt;&lt;/table&gt;&lt;/div&gt;
&lt;h2 id="what-to-do"&gt;What to do &lt;a class="heading-anchor" href="#what-to-do" aria-label="Link to section: What to do"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="for-engineers"&gt;For engineers &lt;a class="heading-anchor" href="#for-engineers" aria-label="Link to section: For engineers"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;On the relying party:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Validate every &lt;a class="term" href="https://www.janbrennenstuhl.eu/glossary/#id-token"&gt;ID token&lt;/a&gt;: signature with a key from the expected issuer&amp;rsquo;s key set, algorithm, issuer, audience, expiry and nonce. Signature manipulation, key confusion, token recipient confusion and replay attacks all exploit a missing check. My essays on &lt;a href="https://www.janbrennenstuhl.eu/jwks/"&gt;JWKS&lt;/a&gt; and &lt;a href="https://www.janbrennenstuhl.eu/jwt-stateless-authentication/"&gt;the purpose of JWT&lt;/a&gt; cover the token side.&lt;/li&gt;
&lt;li&gt;Bind each login to the browser session that started it, and remember which identity provider you sent the user to. Session overwriting and IdP confusion exploit clients that don&amp;rsquo;t.&lt;/li&gt;
&lt;li&gt;Identify users by issuer and subject together, never by subject or email alone.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;On the OpenID provider:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Compare redirect URIs exactly against the registered values.&lt;/li&gt;
&lt;li&gt;Enforce &lt;a class="term" href="https://www.janbrennenstuhl.eu/glossary/#pkce"&gt;&lt;abbr title="Proof Key for Code Exchange"&gt;PKCE&lt;/abbr&gt;&lt;/a&gt; for clients that started with it, so an attacker can&amp;rsquo;t downgrade the flow.&lt;/li&gt;
&lt;li&gt;Accept each authorization code once and only from the client it was issued to.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="for-product-and-ux"&gt;For product and UX &lt;a class="heading-anchor" href="#for-product-and-ux" aria-label="Link to section: For product and UX"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;ul&gt;
&lt;li&gt;Treat &amp;ldquo;OpenID Certified&amp;rdquo; as a conformance signal, not a security audit. Fries tested certified implementations and still disclosed 12 flaws. When you choose a library or vendor, ask for penetration test results that cover these attack classes.&lt;/li&gt;
&lt;li&gt;If your product offers several identity providers, show users which one they&amp;rsquo;re signing in with. IdP confusion attacks rely on the client mixing them up.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="the-studies-in-detail"&gt;The studies in detail &lt;a class="heading-anchor" href="#the-studies-in-detail" aria-label="Link to section: The studies in detail"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="security-analysis-of-real-life-openid-connect-implementations"&gt;Security Analysis of Real-Life OpenID Connect Implementations &lt;a class="heading-anchor" href="#security-analysis-of-real-life-openid-connect-implementations" aria-label="Link to section: Security Analysis of Real-Life OpenID Connect Implementations"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;In this Master&amp;rsquo;s thesis from December 2020, Christian Fries presents a Docker-based lab environment for developers and penetration testers. The lab helps to find security flaws in real-world OpenID Connect Certified service provider and identity provider implementations, and it includes the PrOfESSOS tool for automatic tests. Fries selected 6 identity providers and 8 service providers that support the implicit and hybrid flows. He tested them against 11 service provider attacks and 7 identity provider attacks and disclosed 12 implementation flaws to the developers.&lt;/p&gt;
&lt;p&gt;The thesis also describes common threats and attack scenarios. These include single- and cross-phase attacks on relying parties, like token substitution, key confusion and replay attacks. It also covers attacks on OpenID providers, like message flow confusion, PKCE downgrade and sub claim spoofing.&lt;/p&gt;
&lt;p&gt;&lt;a href="https://www.nds.ruhr-uni-bochum.de/media/ei/arbeiten/2021/05/03/masterthesis.pdf" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="openid-connect-security-considerations"&gt;OpenID Connect Security Considerations &lt;a class="heading-anchor" href="#openid-connect-security-considerations" aria-label="Link to section: OpenID Connect Security Considerations"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;In this 2017 technical report, Vladislav Mladenov and Christian Mainka from Ruhr-University Bochum look at OpenID Connect specification flaws and implementation issues that open up attack vectors. They outline attack scenarios like session overwriting, redirect URI manipulation and token recipient confusion, and give recommendations for defending against them.&lt;/p&gt;
&lt;p&gt;&lt;a href="https://www.nds.ruhr-uni-bochum.de/media/ei/veroeffentlichungen/2017/01/13/OIDCSecurity_1.pdf" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h2 id="related"&gt;Related &lt;a class="heading-anchor" href="#related" aria-label="Link to section: Related"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/step-up-authentication/"&gt;Step-up authentication with OpenID Connect&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-takeover/"&gt;How do attackers take over accounts?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/glossary/"&gt;Glossary&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>How secure are email-based password resets?</title><link>https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-recovery/</link><guid isPermaLink="true">https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-recovery/</guid><pubDate>Fri, 21 Apr 2023 00:00:00 +0200</pubDate><description>&lt;p&gt;&lt;em&gt;Latest change (2026-10-13): Rewritten as a field-guide entry: question title, short answer, key findings and evidence table. Study summary kept and extended with the paper&amp;rsquo;s measured results.&lt;/em&gt;&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; In a 2021 study of 366 websites from the Alexa Top 5K, 72% of account recovery mechanisms had at least one implementation weakness. Common gaps were multiple valid reset tokens, missing change notifications and long-lived reset links.&lt;/p&gt;&lt;h2 id="key-findings"&gt;Key findings &lt;a class="heading-anchor" href="#key-findings" aria-label="Link to section: Key findings"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;figure class="finding" id="f-recovery-72pct-weakness"&gt;
&lt;blockquote&gt;&lt;p&gt;In a study of 366 popular websites, 72% of account recovery mechanisms had at least one implementation weakness.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://kharraz.org/publications/dimva2021youvegotmail.pdf" class="ext"&gt;&lt;cite&gt;Innocenti et al., You&amp;#39;ve Got (a Reset) Mail (DIMVA 2021)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-recovery-72pct-weakness" title="#f-recovery-72pct-weakness"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="In a study of 366 popular websites, 72% of account recovery mechanisms had at least one implementation weakness." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-recovery/#f-recovery-72pct-weakness" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-recovery/f-recovery-72pct-weakness.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-recovery-40pct-multiple-tokens"&gt;
&lt;blockquote&gt;&lt;p&gt;40% of the websites studied generated multiple valid password reset tokens at the same time.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://kharraz.org/publications/dimva2021youvegotmail.pdf" class="ext"&gt;&lt;cite&gt;Innocenti et al., You&amp;#39;ve Got (a Reset) Mail (DIMVA 2021)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-recovery-40pct-multiple-tokens" title="#f-recovery-40pct-multiple-tokens"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="40% of the websites studied generated multiple valid password reset tokens at the same time." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-recovery/#f-recovery-40pct-multiple-tokens" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-recovery/f-recovery-40pct-multiple-tokens.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-recovery-45pct-no-notification"&gt;
&lt;blockquote&gt;&lt;p&gt;45% of the websites studied did not notify the account owner after a password change, so an unauthorised reset went unnoticed.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://kharraz.org/publications/dimva2021youvegotmail.pdf" class="ext"&gt;&lt;cite&gt;Innocenti et al., You&amp;#39;ve Got (a Reset) Mail (DIMVA 2021)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-recovery-45pct-no-notification" title="#f-recovery-45pct-no-notification"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="45% of the websites studied did not notify the account owner after a password change, so an unauthorised reset went unnoticed." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-recovery/#f-recovery-45pct-no-notification" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-recovery/f-recovery-45pct-no-notification.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-recovery-13pct-owasp"&gt;
&lt;blockquote&gt;&lt;p&gt;Only 13% of the websites studied followed the OWASP guidelines for password reset.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://kharraz.org/publications/dimva2021youvegotmail.pdf" class="ext"&gt;&lt;cite&gt;Innocenti et al., You&amp;#39;ve Got (a Reset) Mail (DIMVA 2021)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-recovery-13pct-owasp" title="#f-recovery-13pct-owasp"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Only 13% of the websites studied followed the OWASP guidelines for password reset." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-recovery/#f-recovery-13pct-owasp" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-recovery/f-recovery-13pct-owasp.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;h2 id="evidence"&gt;Evidence &lt;a class="heading-anchor" href="#evidence" aria-label="Link to section: Evidence"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;div class="table-scroll studies" tabindex="0" role="region" aria-label="Account recovery studies"&gt;&lt;table&gt;
&lt;caption&gt;Account recovery studies&lt;/caption&gt;
&lt;thead&gt;&lt;tr&gt;&lt;th scope="col"&gt;Study&lt;/th&gt;&lt;th scope="col"&gt;Year&lt;/th&gt;&lt;th scope="col"&gt;Sample&lt;/th&gt;&lt;th scope="col"&gt;Method&lt;/th&gt;&lt;th scope="col"&gt;Finding&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://kharraz.org/publications/dimva2021youvegotmail.pdf" class="ext"&gt;You&amp;#39;ve Got (a Reset) Mail: A Security Analysis of Email-Based Password Reset Procedures&lt;/a&gt;&lt;span class="org"&gt;Northeastern University, University of Trento, KU Leuven, Florida International University&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2021&lt;/td&gt;&lt;td&gt;366 websites from the Alexa Top 5K&lt;/td&gt;&lt;td&gt;Manual registration and semi-automated checks against 8 OWASP-based controls&lt;/td&gt;&lt;td&gt;72% had at least one weakness; only 13% followed the OWASP guidelines.&lt;/td&gt;&lt;/tr&gt;
&lt;/tbody&gt;&lt;/table&gt;&lt;/div&gt;
&lt;h2 id="what-to-do"&gt;What to do &lt;a class="heading-anchor" href="#what-to-do" aria-label="Link to section: What to do"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="for-engineers"&gt;For engineers &lt;a class="heading-anchor" href="#for-engineers" aria-label="Link to section: For engineers"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;The weaknesses Innocenti et al. measured map to concrete controls:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Keep one valid reset token per account. Invalidate older tokens when a user requests a new one, and accept each token only once.&lt;/li&gt;
&lt;li&gt;Give reset links a short expiry. The study counted any link still valid after 24 hours as a weakness, and found that in 75 websites.&lt;/li&gt;
&lt;li&gt;Send reset links over HTTPS only. 52 websites sent tokens in insecure links.&lt;/li&gt;
&lt;li&gt;End existing sessions after a reset. 38% of the websites kept them alive.&lt;/li&gt;
&lt;li&gt;Protect the flow against login CSRF and header manipulation. Build reset links from configuration, not from the request&amp;rsquo;s host header.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Treat recovery as a high-risk action that deserves the same scrutiny as login. &lt;a href="https://www.janbrennenstuhl.eu/step-up-authentication/"&gt;Step-up authentication&lt;/a&gt; shows how to require more proof for sensitive actions.&lt;/p&gt;
&lt;h3 id="for-product-and-ux"&gt;For product and UX &lt;a class="heading-anchor" href="#for-product-and-ux" aria-label="Link to section: For product and UX"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;ul&gt;
&lt;li&gt;Notify the account owner after every password change, and include a way to report &amp;ldquo;this wasn&amp;rsquo;t me&amp;rdquo;.&lt;/li&gt;
&lt;li&gt;Keep the reset email plain and predictable, so users can spot a fake one.&lt;/li&gt;
&lt;li&gt;Recovery is also an account takeover path. The &lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-takeover/"&gt;account hijacking entry&lt;/a&gt; covers attacks that start before sign-up.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="the-studies-in-detail"&gt;The studies in detail &lt;a class="heading-anchor" href="#the-studies-in-detail" aria-label="Link to section: The studies in detail"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="security-analysis-of-email-based-password-reset-procedures"&gt;Security Analysis of Email-Based Password Reset Procedures &lt;a class="heading-anchor" href="#security-analysis-of-email-based-password-reset-procedures" aria-label="Link to section: Security Analysis of Email-Based Password Reset Procedures"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;An international group of researchers from the USA, Italy and Belgium analyzed the email-based account recovery mechanisms of websites in the Alexa Top 5K. They manually registered accounts and triggered password recovery on sites from 3 popularity groups, ending up with 366 websites. They checked each site against 8 implementation controls derived from the OWASP guidelines.&lt;/p&gt;
&lt;p&gt;The paper describes a heterogeneous reality. 72% of the websites had at least one weakness, and only 13% followed the OWASP guidelines. 40% generated multiple valid tokens, 45% sent no notification after a password change, and 22% were vulnerable to login CSRF. The authors also outline concrete password-recovery attacks that can lead to account takeover.&lt;/p&gt;
&lt;p&gt;&amp;ldquo;We hope that this paper will pave the way in highlighting the importance of improving the email-based account recovery mechanisms in real-world websites.&amp;rdquo;&lt;/p&gt;
&lt;p&gt;&lt;a href="https://kharraz.org/publications/dimva2021youvegotmail.pdf" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h2 id="related"&gt;Related &lt;a class="heading-anchor" href="#related" aria-label="Link to section: Related"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-takeover/"&gt;How do attackers take over accounts?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/reinforced-authentication/"&gt;Do login challenges and risk scoring stop account takeover?&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>How do you protect accounts from stolen credentials?</title><link>https://www.janbrennenstuhl.eu/knowledge-base/account-security/credential-stuffing/</link><guid isPermaLink="true">https://www.janbrennenstuhl.eu/knowledge-base/account-security/credential-stuffing/</guid><pubDate>Fri, 21 Apr 2023 00:00:00 +0200</pubDate><description>&lt;p&gt;&lt;em&gt;Latest change (2026-10-13): Rewritten as a field-guide entry: question title, short answer, key findings and evidence table. Study summaries kept. Replaced the broken link to the password breach alerting paper with the USENIX copy.&lt;/em&gt;&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; Check usernames and passwords against breach data with a privacy-preserving protocol, and add risk signals such as known devices and locations to the login decision. Breached credentials show up in about 1.5% of logins on the web, and in one study 7-25% of exposed passwords matched the victim&amp;rsquo;s Google account.&lt;/p&gt;&lt;h2 id="key-findings"&gt;Key findings &lt;a class="heading-anchor" href="#key-findings" aria-label="Link to section: Key findings"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;figure class="finding" id="f-cred-breached-logins-1-5pct"&gt;
&lt;blockquote&gt;&lt;p&gt;Telemetry from nearly 670,000 users and 21 million logins showed that 1.5% of logins on the web involve breached credentials.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.usenix.org/system/files/sec19-thomas.pdf" class="ext"&gt;&lt;cite&gt;Thomas et al., Protecting accounts from credential stuffing with password breach alerting (USENIX Security 2019)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-cred-breached-logins-1-5pct" title="#f-cred-breached-logins-1-5pct"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Telemetry from nearly 670,000 users and 21 million logins showed that 1.5% of logins on the web involve breached credentials." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/credential-stuffing/#f-cred-breached-logins-1-5pct" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/credential-stuffing/f-cred-breached-logins-1-5pct.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-cred-alert-26pct-new-password"&gt;
&lt;blockquote&gt;&lt;p&gt;26% of breach warnings shown at login led users to switch to a new password at least as strong as the original.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.usenix.org/system/files/sec19-thomas.pdf" class="ext"&gt;&lt;cite&gt;Thomas et al., Protecting accounts from credential stuffing with password breach alerting (USENIX Security 2019)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-cred-alert-26pct-new-password" title="#f-cred-alert-26pct-new-password"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="26% of breach warnings shown at login led users to switch to a new password at least as strong as the original." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/credential-stuffing/#f-cred-alert-26pct-new-password" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/credential-stuffing/f-cred-alert-26pct-new-password.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-cred-google-match-7-25pct"&gt;
&lt;blockquote&gt;&lt;p&gt;Between 7% and 25% of passwords exposed through breaches, phishing kits and keyloggers matched the victim&amp;rsquo;s Google account.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://dl.acm.org/doi/pdf/10.1145/3133956.3134067" class="ext"&gt;&lt;cite&gt;Thomas et al., Data Breaches, Phishing, or Malware? (CCS 2017)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-cred-google-match-7-25pct" title="#f-cred-google-match-7-25pct"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Between 7% and 25% of passwords exposed through breaches, phishing kits and keyloggers matched the victim&amp;rsquo;s Google account." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/credential-stuffing/#f-cred-google-match-7-25pct" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/credential-stuffing/f-cred-google-match-7-25pct.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-cred-hash-prefix-12x"&gt;
&lt;blockquote&gt;&lt;p&gt;In some contexts, the hash prefixes that common breach-checking services receive make remote guessing attacks 12 times more effective.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://rist.tech.cornell.edu/papers/c3.pdf" class="ext"&gt;&lt;cite&gt;Li et al., Protocols for Checking Compromised Credentials (CCS 2019)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-cred-hash-prefix-12x" title="#f-cred-hash-prefix-12x"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="In some contexts, the hash prefixes that common breach-checking services receive make remote guessing attacks 12 times more effective." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/credential-stuffing/#f-cred-hash-prefix-12x" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/credential-stuffing/f-cred-hash-prefix-12x.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;h2 id="evidence"&gt;Evidence &lt;a class="heading-anchor" href="#evidence" aria-label="Link to section: Evidence"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;div class="table-scroll studies" tabindex="0" role="region" aria-label="Compromised credential studies"&gt;&lt;table&gt;
&lt;caption&gt;Compromised credential studies&lt;/caption&gt;
&lt;thead&gt;&lt;tr&gt;&lt;th scope="col"&gt;Study&lt;/th&gt;&lt;th scope="col"&gt;Year&lt;/th&gt;&lt;th scope="col"&gt;Sample&lt;/th&gt;&lt;th scope="col"&gt;Method&lt;/th&gt;&lt;th scope="col"&gt;Finding&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://rist.tech.cornell.edu/papers/c3.pdf" class="ext"&gt;Protocols for Checking Compromised Credentials&lt;/a&gt;&lt;span class="org"&gt;Cornell Tech, Cloudflare&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2019&lt;/td&gt;&lt;td&gt;Existing C3 services (HaveIBeenPwned, Google Password Checkup)&lt;/td&gt;&lt;td&gt;Formal analysis, leakage simulation, protocol design&lt;/td&gt;&lt;td&gt;Hash-prefix leakage can make guessing 12 times more effective; 2 stronger protocols remain practical.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.usenix.org/system/files/sec19-thomas.pdf" class="ext"&gt;Protecting accounts from credential stuffing with password breach alerting&lt;/a&gt;&lt;span class="org"&gt;Google, Stanford&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2019&lt;/td&gt;&lt;td&gt;Over 4 billion breached credentials; nearly 670,000 users and 21 million logins&lt;/td&gt;&lt;td&gt;Protocol design, Chrome extension telemetry&lt;/td&gt;&lt;td&gt;1.5% of logins involve breached credentials; 26% of warnings led to a new password.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://dl.acm.org/doi/pdf/10.1145/3133956.3134067" class="ext"&gt;Data Breaches, Phishing, or Malware? Understanding the Risks of Stolen Credentials&lt;/a&gt;&lt;span class="org"&gt;Google, UC Berkeley, ICSI&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2017&lt;/td&gt;&lt;td&gt;1.9 billion breached credentials, 12.4 million phishing and 788,000 keylogger victims&lt;/td&gt;&lt;td&gt;Longitudinal measurement, March 2016 to March 2017&lt;/td&gt;&lt;td&gt;7-25% of exposed passwords matched the victim&amp;rsquo;s Google account.&lt;/td&gt;&lt;/tr&gt;
&lt;/tbody&gt;&lt;/table&gt;&lt;/div&gt;
&lt;h2 id="what-to-do"&gt;What to do &lt;a class="heading-anchor" href="#what-to-do" aria-label="Link to section: What to do"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="for-engineers"&gt;For engineers &lt;a class="heading-anchor" href="#for-engineers" aria-label="Link to section: For engineers"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;ul&gt;
&lt;li&gt;Check credentials against breach data at sign-up, password change and login. Use a protocol that never sends the password or its full hash to the checking service. Thomas et al. combine &amp;ldquo;computationally expensive hashing, k-anonymity, and private set intersection&amp;rdquo;; Li et al. show how much a plain hash prefix can leak.&lt;/li&gt;
&lt;li&gt;Don&amp;rsquo;t decide a login on the password alone. Thomas et al. (2017) show that risk signals such as a user&amp;rsquo;s historical geolocations and device profiles reduce the hijacking risk from stolen passwords. The &lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/reinforced-authentication/"&gt;reinforced authentication entry&lt;/a&gt; collects studies on login risk scoring.&lt;/li&gt;
&lt;li&gt;Watch for stuffing and spraying patterns across accounts, not only per account. See &lt;a href="https://www.janbrennenstuhl.eu/password-spraying/"&gt;password spraying&lt;/a&gt; and &lt;a href="https://www.janbrennenstuhl.eu/honeytoken/"&gt;honeytokens&lt;/a&gt; for detection ideas.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="for-product-and-ux"&gt;For product and UX &lt;a class="heading-anchor" href="#for-product-and-ux" aria-label="Link to section: For product and UX"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;ul&gt;
&lt;li&gt;Tell users when their credentials appear in a breach, and make the next step a single action. In the Google study, 26% of warnings led to a new password that was at least as strong.&lt;/li&gt;
&lt;li&gt;Expect most users not to act on a warning alone. The &lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-reuse/"&gt;password reuse entry&lt;/a&gt; covers what makes a notification work and where notifications fall short.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="the-studies-in-detail"&gt;The studies in detail &lt;a class="heading-anchor" href="#the-studies-in-detail" aria-label="Link to section: The studies in detail"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="protocols-for-checking-compromised-credentials"&gt;Protocols for Checking Compromised Credentials &lt;a class="heading-anchor" href="#protocols-for-checking-compromised-credentials" aria-label="Link to section: Protocols for Checking Compromised Credentials"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;&amp;ldquo;To prevent credential stuffing attacks, industry best practice now proactively checks if user credentials are present in known data breaches.&amp;rdquo;&lt;/p&gt;
&lt;p&gt;For this paper, researchers from Cornell Tech and Cloudflare analyzed existing compromised credential checking (C3) services such as HaveIBeenPwned and Google Password Checkup. They give the first formal description of C3 services, including operational requirements and threat models. Current services have the client share a short prefix of a password hash. The authors show that in some contexts this prefix makes remote guessing 12 times more effective. They then propose &amp;ldquo;two new protocols that provide stronger protection for users&amp;rsquo; passwords&amp;rdquo;, implement them, and show that they remain practical to deploy.&lt;/p&gt;
&lt;p&gt;&lt;a href="https://rist.tech.cornell.edu/papers/c3.pdf" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="protecting-accounts-from-credential-stuffing-with-password-breach-alerting"&gt;Protecting accounts from credential stuffing with password breach alerting &lt;a class="heading-anchor" href="#protecting-accounts-from-credential-stuffing-with-password-breach-alerting" aria-label="Link to section: Protecting accounts from credential stuffing with password breach alerting"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;Researchers from Google and Stanford describe how Google&amp;rsquo;s password breach alerting works, which constraints and hard requirements it meets, and which cryptographic protocols it uses. Their cloud service covered over 4 billion breached credentials, and a Chrome extension served as the first client.&lt;/p&gt;
&lt;blockquote&gt;
&lt;p&gt;&amp;ldquo;Our protocol relies on a combination of computationally expensive hashing, k-anonymity, and private set intersection. Our approach improves on existing protocols by taking into account both an adversarial client and server, while also minimizing the chance of false positives.&amp;rdquo;&lt;/p&gt;
&lt;/blockquote&gt;
&lt;p&gt;Anonymous telemetry from nearly 670,000 users and 21 million logins showed that 1.5% of logins on the web involved breached credentials. 26% of the warnings led users to migrate to a new password at least as strong as the original.&lt;/p&gt;
&lt;p&gt;&lt;a href="https://www.usenix.org/system/files/sec19-thomas.pdf" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="data-breaches-phishing-or-malware-understanding-the-risks-of-stolen-credentials"&gt;Data Breaches, Phishing, or Malware? Understanding the Risks of Stolen Credentials &lt;a class="heading-anchor" href="#data-breaches-phishing-or-malware-understanding-the-risks-of-stolen-credentials" aria-label="Link to section: Data Breaches, Phishing, or Malware? Understanding the Risks of Stolen Credentials"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;Between March 2016 and March 2017, researchers from Google, UC Berkeley and the International Computer Science Institute identified 788,000 potential keylogger victims, 12.4 million potential phishing kit victims and 1.9 billion usernames and passwords exposed in data breaches. With this data, they &amp;ldquo;explore to what degree the stolen passwords […] enable an attacker to obtain a victim&amp;rsquo;s valid email credentials — and thus complete control of their online identity due to transitive trust&amp;rdquo;. In their case study of Google accounts, 7-25% of exposed passwords matched the victim&amp;rsquo;s Google account.&lt;/p&gt;
&lt;p&gt;The study is from 2017, but its mitigation still applies today: the authors show that adding risk signals, such as a user&amp;rsquo;s historical geolocations and device profiles, to authentication helps mitigate hijacking with a correct but stolen password.&lt;/p&gt;
&lt;p&gt;&lt;a href="https://dl.acm.org/doi/pdf/10.1145/3133956.3134067" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h2 id="related"&gt;Related &lt;a class="heading-anchor" href="#related" aria-label="Link to section: Related"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-reuse/"&gt;How common is password reuse?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/reinforced-authentication/"&gt;Do login challenges and risk scoring stop account takeover?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/step-up-authentication/"&gt;Step-up authentication&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>How common is password reuse?</title><link>https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-reuse/</link><guid isPermaLink="true">https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-reuse/</guid><pubDate>Thu, 27 Apr 2023 00:00:00 +0200</pubDate><description>&lt;p&gt;&lt;em&gt;Latest change (2026-10-13): Rewritten as a field-guide entry: question title, short answer, key findings and evidence table. Study summaries kept. Added sample sizes and numbers from the papers&amp;rsquo; own text and corrected the affiliations for The Tangled Web of Password Reuse.&lt;/em&gt;&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; Password reuse is very common. One study estimates that 43-51% of users reuse the same password across sites, and an in situ study found exact or partial reuse in about 60% of unique passwords. An attacker who holds a user&amp;rsquo;s password from one site can guess that user&amp;rsquo;s password elsewhere far more easily.&lt;/p&gt;&lt;h2 id="key-findings"&gt;Key findings &lt;a class="heading-anchor" href="#key-findings" aria-label="Link to section: Key findings"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;figure class="finding" id="f-pw-reuse-43-51pct"&gt;
&lt;blockquote&gt;&lt;p&gt;An estimated 43-51% of users reuse the same password across multiple sites.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.ndss-symposium.org/wp-content/uploads/2017/09/06_1_1.pdf" class="ext"&gt;&lt;cite&gt;Das et al., The Tangled Web of Password Reuse (NDSS 2014)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-pw-reuse-43-51pct" title="#f-pw-reuse-43-51pct"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="An estimated 43-51% of users reuse the same password across multiple sites." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-reuse/#f-pw-reuse-43-51pct" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-reuse/f-pw-reuse-43-51pct.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-pw-reuse-cross-site-guessing"&gt;
&lt;blockquote&gt;&lt;p&gt;With a user&amp;rsquo;s password from another site, a cross-site guessing algorithm guessed 30% of transformed passwords within 100 attempts, compared with 14% for a standard guessing algorithm.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.ndss-symposium.org/wp-content/uploads/2017/09/06_1_1.pdf" class="ext"&gt;&lt;cite&gt;Das et al., The Tangled Web of Password Reuse (NDSS 2014)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-pw-reuse-cross-site-guessing" title="#f-pw-reuse-cross-site-guessing"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="With a user&amp;rsquo;s password from another site, a cross-site guessing algorithm guessed 30% of transformed passwords within 100 attempts, compared with 14% for a standard guessing algorithm." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-reuse/#f-pw-reuse-cross-site-guessing" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-reuse/f-pw-reuse-cross-site-guessing.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-pw-reuse-partial-60pct"&gt;
&lt;blockquote&gt;&lt;p&gt;When researchers observed 154 people&amp;rsquo;s real logins for an average of 147 days, about 60% of their unique passwords were exactly or partially reused.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://dl.acm.org/doi/pdf/10.1145/3133956.3133973" class="ext"&gt;&lt;cite&gt;Pearman et al., Let&amp;#39;s Go in for a Closer Look: Observing Passwords in Their Natural Habitat (CCS 2017)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-pw-reuse-partial-60pct" title="#f-pw-reuse-partial-60pct"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="When researchers observed 154 people&amp;rsquo;s real logins for an average of 147 days, about 60% of their unique passwords were exactly or partially reused." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-reuse/#f-pw-reuse-partial-60pct" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-reuse/f-pw-reuse-partial-60pct.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-pw-reuse-frequent-entry"&gt;
&lt;blockquote&gt;&lt;p&gt;People reuse each password on 1.7-3.4 websites, and they mostly reuse the complex passwords they have to enter frequently.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.usenix.org/system/files/conference/soups2016/soups2016-paper-wash.pdf" class="ext"&gt;&lt;cite&gt;Wash et al., Understanding Password Choices (SOUPS 2016)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-pw-reuse-frequent-entry" title="#f-pw-reuse-frequent-entry"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="People reuse each password on 1.7-3.4 websites, and they mostly reuse the complex passwords they have to enter frequently." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-reuse/#f-pw-reuse-frequent-entry" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-reuse/f-pw-reuse-frequent-entry.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-pw-reuse-notifications-third"&gt;
&lt;blockquote&gt;&lt;p&gt;Fewer than a third of people shown a typical password-reuse notification said they intended to change any passwords.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://dl.acm.org/doi/pdf/10.1145/3243734.3243767" class="ext"&gt;&lt;cite&gt;Golla et al., Designing Password-Reuse Notifications (CCS 2018)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-pw-reuse-notifications-third" title="#f-pw-reuse-notifications-third"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Fewer than a third of people shown a typical password-reuse notification said they intended to change any passwords." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-reuse/#f-pw-reuse-notifications-third" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-reuse/f-pw-reuse-notifications-third.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;h2 id="evidence"&gt;Evidence &lt;a class="heading-anchor" href="#evidence" aria-label="Link to section: Evidence"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;div class="table-scroll studies" tabindex="0" role="region" aria-label="Password reuse studies"&gt;&lt;table&gt;
&lt;caption&gt;Password reuse studies&lt;/caption&gt;
&lt;thead&gt;&lt;tr&gt;&lt;th scope="col"&gt;Study&lt;/th&gt;&lt;th scope="col"&gt;Year&lt;/th&gt;&lt;th scope="col"&gt;Sample&lt;/th&gt;&lt;th scope="col"&gt;Method&lt;/th&gt;&lt;th scope="col"&gt;Finding&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://dl.acm.org/doi/pdf/10.1145/3133956.3133973" class="ext"&gt;Let&amp;#39;s Go in for a Closer Look: Observing Passwords in Their Natural Habitat&lt;/a&gt;&lt;span class="org"&gt;Carnegie Mellon University, ICSI, Google&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2017&lt;/td&gt;&lt;td&gt;154 participants, 147 days on average&lt;/td&gt;&lt;td&gt;In situ measurement on participants&amp;#39; computers&lt;/td&gt;&lt;td&gt;About 60% of unique passwords were exactly or partially reused.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.ndss-symposium.org/wp-content/uploads/2017/09/06_1_1.pdf" class="ext"&gt;The Tangled Web of Password Reuse&lt;/a&gt;&lt;span class="org"&gt;University of Illinois, Princeton, Indiana University&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2014&lt;/td&gt;&lt;td&gt;Several hundred thousand leaked passwords from 11 sites; 224 survey responses&lt;/td&gt;&lt;td&gt;Leak analysis, survey, guessing experiment&lt;/td&gt;&lt;td&gt;Estimated 43-51% of users reuse the same password across sites.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.usenix.org/system/files/conference/soups2016/soups2016-paper-wash.pdf" class="ext"&gt;Understanding Password Choices: How Frequently Entered Passwords are Re-used Across Websites&lt;/a&gt;&lt;span class="org"&gt;Michigan State University, Macalester College&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2016&lt;/td&gt;&lt;td&gt;134 participants over 6 weeks&lt;/td&gt;&lt;td&gt;Survey combined with measured online behaviour&lt;/td&gt;&lt;td&gt;Each password is reused on 1.7-3.4 websites, mostly the frequently entered ones.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.microsoft.com/en-us/research/wp-content/uploads/2006/11/www2007.pdf" class="ext"&gt;A Large-Scale Study of Web Password Habits&lt;/a&gt;&lt;span class="org"&gt;Microsoft Research&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2007&lt;/td&gt;&lt;td&gt;Over half a million users, 3 months&lt;/td&gt;&lt;td&gt;Client-side measurement&lt;/td&gt;&lt;td&gt;The average user had 6.5 passwords, each shared across 3.9 sites.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://dl.acm.org/doi/pdf/10.1145/3243734.3243767" class="ext"&gt;Designing Password-Reuse Notifications&lt;/a&gt;&lt;/th&gt;&lt;td&gt;2018&lt;/td&gt;&lt;td&gt;180 and 588 respondents&lt;/td&gt;&lt;td&gt;2 online user studies of notification designs&lt;/td&gt;&lt;td&gt;Fewer than a third of respondents in study 1 intended to change any password.&lt;/td&gt;&lt;/tr&gt;
&lt;/tbody&gt;&lt;/table&gt;&lt;/div&gt;
&lt;h2 id="what-to-do"&gt;What to do &lt;a class="heading-anchor" href="#what-to-do" aria-label="Link to section: What to do"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="for-engineers"&gt;For engineers &lt;a class="heading-anchor" href="#for-engineers" aria-label="Link to section: For engineers"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;ul&gt;
&lt;li&gt;Assume that some of your users&amp;rsquo; passwords are already in another site&amp;rsquo;s breach. Check passwords against breach data at sign-up, password change and login. The &lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/credential-stuffing/"&gt;compromised credentials entry&lt;/a&gt; covers protocols that do this without sending the password to a third party.&lt;/li&gt;
&lt;li&gt;Treat a correct password as one signal among several. Das et al. show that a leaked password from another site gives an online attacker a large head start, so rate limits alone aren&amp;rsquo;t enough. See &lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/reinforced-authentication/"&gt;reinforced authentication&lt;/a&gt; and &lt;a href="https://www.janbrennenstuhl.eu/step-up-authentication/"&gt;step-up authentication&lt;/a&gt;.&lt;/li&gt;
&lt;li&gt;Make unique passwords cheap for users: allow paste, long passwords and autofill so a password manager works. I wrote about this in &lt;a href="https://www.janbrennenstuhl.eu/password-manager-friendly/"&gt;password-manager-friendly login forms&lt;/a&gt; and &lt;a href="https://www.janbrennenstuhl.eu/long-password-support/"&gt;long password support&lt;/a&gt;.&lt;/li&gt;
&lt;li&gt;Where you can, offer a factor that can&amp;rsquo;t be reused at all, such as passkeys. See my &lt;a href="https://www.janbrennenstuhl.eu/fido-passkeys/"&gt;critical notes on passkeys&lt;/a&gt;.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="for-product-and-ux"&gt;For product and UX &lt;a class="heading-anchor" href="#for-product-and-ux" aria-label="Link to section: For product and UX"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;ul&gt;
&lt;li&gt;Composition rules can backfire. Wash et al. found that people memorize a strong password for a site they log into often and then reuse it elsewhere. Read &lt;a href="https://www.janbrennenstuhl.eu/password-policy/"&gt;password policy&lt;/a&gt; before adding another character-class rule.&lt;/li&gt;
&lt;li&gt;Don&amp;rsquo;t rely on a notification alone. Golla et al. conclude that &amp;ldquo;notifications alone appear insufficient in solving password reuse&amp;rdquo;. Pair the message with a forced reset and a clear next step, such as using a password manager.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="the-studies-in-detail"&gt;The studies in detail &lt;a class="heading-anchor" href="#the-studies-in-detail" aria-label="Link to section: The studies in detail"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="observing-passwords-in-their-natural-habitat"&gt;Observing Passwords in Their Natural Habitat &lt;a class="heading-anchor" href="#observing-passwords-in-their-natural-habitat" aria-label="Link to section: Observing Passwords in Their Natural Habitat"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;Researchers from Carnegie Mellon University, the International Computer Science Institute and Google instrumented the computers of 154 participants for an average of 147 days each. They wanted to analyze &amp;ldquo;password characteristics and behaviors across the full range of participants&amp;rsquo; web-based accounts&amp;rdquo;.&lt;/p&gt;
&lt;p&gt;Among other things, the data showed &amp;ldquo;that the use of symbols and digits in passwords predicts increased likelihood of reuse, while increased password strength predicts decreased likelihood of reuse; that password reuse is more prevalent than previously believed, especially when partial reuse is taken into account&amp;rdquo;. Of 1,578 unique passwords, 32% were exactly reused and about 60% showed some form of reuse.&lt;/p&gt;
&lt;p&gt;&lt;a href="https://dl.acm.org/doi/pdf/10.1145/3133956.3133973" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="the-tangled-web-of-password-reuse"&gt;The Tangled Web of Password Reuse &lt;a class="heading-anchor" href="#the-tangled-web-of-password-reuse" aria-label="Link to section: The Tangled Web of Password Reuse"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;This 2014 study by researchers from the University of Illinois, Princeton and Indiana University used several hundred thousand leaked passwords from 11 websites and a user survey with 224 responses. The authors estimate that &amp;ldquo;43-51% of users reuse the same password across multiple sites&amp;rdquo;. They also identified &amp;ldquo;a few simple tricks users often employ to transform a basic password between sites&amp;rdquo; and built the first cross-site password-guessing algorithm.&lt;/p&gt;
&lt;p&gt;That algorithm guessed 30% of transformed passwords within 100 attempts, compared with 14% without knowledge of the other site&amp;rsquo;s password. A leaked password helps the attacker even when the user changed it slightly for the next site.&lt;/p&gt;
&lt;p&gt;&lt;a href="https://www.ndss-symposium.org/wp-content/uploads/2017/09/06_1_1.pdf" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="understanding-password-choices-how-frequently-entered-passwords-are-re-used-across-websites"&gt;Understanding Password Choices: How Frequently Entered Passwords are Re-used Across Websites &lt;a class="heading-anchor" href="#understanding-password-choices-how-frequently-entered-passwords-are-re-used-across-websites" aria-label="Link to section: Understanding Password Choices: How Frequently Entered Passwords are Re-used Across Websites"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;Researchers from Michigan State University and Macalester College combined survey answers with the measured online behavior of 134 participants over 6 weeks. They found &amp;ldquo;that people do tend to re-use each password on 1.7–3.4 different websites&amp;rdquo; and that &amp;ldquo;they reuse passwords that are more complex, and mostly they tend to re-use passwords that they have to enter frequently&amp;rdquo;.&lt;/p&gt;
&lt;p&gt;The most interesting observation is how accounts depend on each other, which shows the trouble with arbitrary password composition requirements:&lt;/p&gt;
&lt;blockquote&gt;
&lt;p&gt;&amp;ldquo;[…] if users must memorize a strong password on a website where they have to enter it frequently, they then re-use it elsewhere. This results in stronger passwords on more websites. This practice puts users at greater risk of cross-site password guessing attacks […]&amp;rdquo;&lt;/p&gt;
&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://www.usenix.org/system/files/conference/soups2016/soups2016-paper-wash.pdf" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="a-large-scale-study-of-web-password-habits"&gt;A Large-Scale Study of Web Password Habits &lt;a class="heading-anchor" href="#a-large-scale-study-of-web-password-habits" aria-label="Link to section: A Large-Scale Study of Web Password Habits"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;Microsoft researchers measured the password use and reuse of over half a million users over 3 months in 2007, using a client component on users&amp;rsquo; machines. It was the first large-scale study of its kind. The average user had 6.5 passwords, each shared across 3.9 different sites, and about 25 accounts that required a password. The authors write that the results &amp;ldquo;particularly confirm the conventional wisdom about the large number and poor quality of user passwords. In addition passwords are re-used and forgotten a great deal.&amp;rdquo;&lt;/p&gt;
&lt;p&gt;&lt;a href="https://www.microsoft.com/en-us/research/wp-content/uploads/2006/11/www2007.pdf" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="designing-password-reuse-notifications"&gt;Designing Password-Reuse Notifications &lt;a class="heading-anchor" href="#designing-password-reuse-notifications" aria-label="Link to section: Designing Password-Reuse Notifications"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;When companies find stolen credentials and notice password reuse, they may require a password reset and notify affected users. Researchers from Germany and the USA ran 2 user studies in 2018 to learn how people react to such notifications. In the first, 180 respondents each saw 1 of 6 notifications that real companies had used, and fewer than a third reported an intention to change any passwords. In the second, 588 respondents saw 1 of 15 variations of a model notification. The authors discuss &amp;ldquo;best practices for password-reuse notifications and how notifications alone appear insufficient in solving password reuse&amp;rdquo;.&lt;/p&gt;
&lt;p&gt;&lt;a href="https://dl.acm.org/doi/pdf/10.1145/3243734.3243767" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h2 id="related"&gt;Related &lt;a class="heading-anchor" href="#related" aria-label="Link to section: Related"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/credential-stuffing/"&gt;How do you protect accounts from stolen credentials?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-strength-meter/"&gt;Do password strength meters work?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/password-spraying/"&gt;Password spraying&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/max-password-length/"&gt;How long a password should be allowed to be&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>Do password strength meters work?</title><link>https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-strength-meter/</link><guid isPermaLink="true">https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-strength-meter/</guid><pubDate>Wed, 03 May 2023 00:00:00 +0200</pubDate><description>&lt;p&gt;&lt;em&gt;Latest change (2026-10-13): Rewritten as a field-guide entry: question title, short answer, key findings and evidence table. Study summaries kept. Corrected the summary of the meter experiment and the zxcvbn guess count (10^5, not 105).&lt;/em&gt;&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; Meters work under the right conditions. In one study they produced stronger passwords when people changed passwords on important accounts, and made no observable difference for an unimportant one. Many deployed meters also label weak passwords as strong, so a useful meter has to estimate how guessable a password really is.&lt;/p&gt;&lt;h2 id="key-findings"&gt;Key findings &lt;a class="heading-anchor" href="#key-findings" aria-label="Link to section: Key findings"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;figure class="finding" id="f-meter-important-accounts"&gt;
&lt;blockquote&gt;&lt;p&gt;Password meters produced significantly stronger passwords when people had to change their real passwords on important accounts.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/chi13b.pdf" class="ext"&gt;&lt;cite&gt;Egelman et al., Does My Password Go up to Eleven? (CHI 2013)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-meter-important-accounts" title="#f-meter-important-accounts"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Password meters produced significantly stronger passwords when people had to change their real passwords on important accounts." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-strength-meter/#f-meter-important-accounts" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-strength-meter/f-meter-important-accounts.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-meter-unimportant-no-effect"&gt;
&lt;blockquote&gt;&lt;p&gt;When people created a password for an unimportant account, meters made no observable difference because they reused weak passwords from similar low-risk accounts.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/chi13b.pdf" class="ext"&gt;&lt;cite&gt;Egelman et al., Does My Password Go up to Eleven? (CHI 2013)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-meter-unimportant-no-effect" title="#f-meter-unimportant-no-effect"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="When people created a password for an unimportant account, meters made no observable difference because they reused weak passwords from similar low-risk accounts." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-strength-meter/#f-meter-unimportant-no-effect" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-strength-meter/f-meter-unimportant-no-effect.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-meter-inconsistent-labels"&gt;
&lt;blockquote&gt;&lt;p&gt;Password meters at popular web services gave highly inconsistent ratings for the same password and labeled many weak passwords as strong or even very strong.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://xavier2dc.fr/papers/password-meters-ndss2014.pdf" class="ext"&gt;&lt;cite&gt;de Carné de Carnavalet and Mannan, From Very Weak to Very Strong: Analyzing Password-Strength Meters (NDSS 2014)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-meter-inconsistent-labels" title="#f-meter-inconsistent-labels"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Password meters at popular web services gave highly inconsistent ratings for the same password and labeled many weak passwords as strong or even very strong." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-strength-meter/#f-meter-inconsistent-labels" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-strength-meter/f-meter-inconsistent-labels.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-meter-zxcvbn-1-5mb"&gt;
&lt;blockquote&gt;&lt;p&gt;A strength estimator with 1.5 MB of compressed data can accurately estimate the best-known guessing attacks up to 10^5 guesses.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.usenix.org/system/files/conference/usenixsecurity16/sec16_paper_wheeler.pdf" class="ext"&gt;&lt;cite&gt;Wheeler, zxcvbn: Low-Budget Password Strength Estimation (USENIX Security 2016)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-meter-zxcvbn-1-5mb" title="#f-meter-zxcvbn-1-5mb"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="A strength estimator with 1.5 MB of compressed data can accurately estimate the best-known guessing attacks up to 10^5 guesses." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-strength-meter/#f-meter-zxcvbn-1-5mb" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-strength-meter/f-meter-zxcvbn-1-5mb.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-meter-policy-strength-length"&gt;
&lt;blockquote&gt;&lt;p&gt;Researchers recommend combining minimum-strength and minimum-length password requirements for high-value accounts, and found that adding blocklists provides no further gains.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://dl.acm.org/doi/pdf/10.1145/3372297.3417882" class="ext"&gt;&lt;cite&gt;Tan et al., Practical Recommendations for Stronger, More Usable Passwords (CCS 2020)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-meter-policy-strength-length" title="#f-meter-policy-strength-length"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Researchers recommend combining minimum-strength and minimum-length password requirements for high-value accounts, and found that adding blocklists provides no further gains." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-strength-meter/#f-meter-policy-strength-length" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-strength-meter/f-meter-policy-strength-length.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;h2 id="evidence"&gt;Evidence &lt;a class="heading-anchor" href="#evidence" aria-label="Link to section: Evidence"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;div class="table-scroll studies" tabindex="0" role="region" aria-label="Password strength meter studies"&gt;&lt;table&gt;
&lt;caption&gt;Password strength meter studies&lt;/caption&gt;
&lt;thead&gt;&lt;tr&gt;&lt;th scope="col"&gt;Study&lt;/th&gt;&lt;th scope="col"&gt;Year&lt;/th&gt;&lt;th scope="col"&gt;Sample&lt;/th&gt;&lt;th scope="col"&gt;Method&lt;/th&gt;&lt;th scope="col"&gt;Finding&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/chi13b.pdf" class="ext"&gt;Does My Password Go up to Eleven? The Impact of Password Meters on Password Selection&lt;/a&gt;&lt;span class="org"&gt;UC Berkeley, University of British Columbia, Microsoft Research&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2013&lt;/td&gt;&lt;td&gt;47 lab participants; 541 field participants&lt;/td&gt;&lt;td&gt;Laboratory experiment with real password changes, field experiment&lt;/td&gt;&lt;td&gt;Meters helped on important accounts and made no observable difference on an unimportant one.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://xavier2dc.fr/papers/password-meters-ndss2014.pdf" class="ext"&gt;From Very Weak to Very Strong: Analyzing Password-Strength Meters&lt;/a&gt;&lt;span class="org"&gt;Concordia University&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2014&lt;/td&gt;&lt;td&gt;13 meters from 11 web services; nearly 4 million passwords&lt;/td&gt;&lt;td&gt;Black-box testing with password dictionaries&lt;/td&gt;&lt;td&gt;Meters disagree with each other and rate many weak passwords as strong.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.usenix.org/system/files/conference/usenixsecurity16/sec16_paper_wheeler.pdf" class="ext"&gt;zxcvbn: Low-Budget Password Strength Estimation&lt;/a&gt;&lt;span class="org"&gt;Dropbox&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2016&lt;/td&gt;&lt;td&gt;Leaked password sets&lt;/td&gt;&lt;td&gt;Comparison against 4 modern guessing attacks&lt;/td&gt;&lt;td&gt;1.5 MB is enough to estimate guessing attacks accurately up to 10^5 guesses.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.usenix.org/system/files/conference/usenixsecurity16/sec16_paper_melicher.pdf" class="ext"&gt;Fast, Lean, and Accurate: Modeling Password Guessability Using Neural Networks&lt;/a&gt;&lt;span class="org"&gt;Carnegie Mellon University&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2016&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Neural network guessing models, client-side JavaScript implementation&lt;/td&gt;&lt;td&gt;Neural networks often guess passwords better than earlier models and compress to hundreds of kilobytes.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://dl.acm.org/doi/pdf/10.1145/3372297.3417882" class="ext"&gt;Practical Recommendations for Stronger, More Usable Passwords&lt;/a&gt;&lt;span class="org"&gt;Carnegie Mellon University&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2020&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;2 online experiments&lt;/td&gt;&lt;td&gt;Combine minimum-strength and minimum-length requirements; blocklists add no further gains.&lt;/td&gt;&lt;/tr&gt;
&lt;/tbody&gt;&lt;/table&gt;&lt;/div&gt;
&lt;h2 id="what-to-do"&gt;What to do &lt;a class="heading-anchor" href="#what-to-do" aria-label="Link to section: What to do"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="for-engineers"&gt;For engineers &lt;a class="heading-anchor" href="#for-engineers" aria-label="Link to section: For engineers"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;ul&gt;
&lt;li&gt;Estimate strength from guessability, using a model such as zxcvbn or a neural network, and stop counting character classes. Wheeler calls counts of lower- and uppercase letters, digits and symbols &amp;ldquo;a conclusively burdensome and ineffective security practice&amp;rdquo;.&lt;/li&gt;
&lt;li&gt;For high-value accounts, combine a minimum length with a minimum-strength check. Tan et al. found that character-class requirements may provide very little improvement against expert attackers &amp;ldquo;and may even reduce effective security&amp;rdquo;.&lt;/li&gt;
&lt;li&gt;Allow long passwords and passphrases so the strength check has something to reward. See &lt;a href="https://www.janbrennenstuhl.eu/max-password-length/"&gt;password length&lt;/a&gt; and &lt;a href="https://www.janbrennenstuhl.eu/long-password-support/"&gt;long password support&lt;/a&gt;.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="for-product-and-ux"&gt;For product and UX &lt;a class="heading-anchor" href="#for-product-and-ux" aria-label="Link to section: For product and UX"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;ul&gt;
&lt;li&gt;Never show &amp;ldquo;strong&amp;rdquo; for a weak password. de Carné de Carnavalet and Mannan conclude that meters &amp;ldquo;should avoid providing misleading strength outcomes, especially for weak passwords&amp;rdquo;.&lt;/li&gt;
&lt;li&gt;Give advice people can act on, not only a coloured bar. Melicher et al. name &amp;ldquo;giving user-interpretable advice to improve passwords during password creation&amp;rdquo; as the remaining challenge.&lt;/li&gt;
&lt;li&gt;Put effort into meters where they matter: password changes on important accounts. On throwaway sign-ups, a meter won&amp;rsquo;t stop people from reusing a weak password. My &lt;a href="https://www.janbrennenstuhl.eu/password-policy/"&gt;password policy guide&lt;/a&gt; covers the rest of the policy.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="the-studies-in-detail"&gt;The studies in detail &lt;a class="heading-anchor" href="#the-studies-in-detail" aria-label="Link to section: The studies in detail"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="impact-of-password-meters-on-password-selection"&gt;Impact of Password Meters on Password Selection &lt;a class="heading-anchor" href="#impact-of-password-meters-on-password-selection" aria-label="Link to section: Impact of Password Meters on Password Selection"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;Researchers from UC Berkeley, the University of British Columbia and Microsoft Research performed &amp;ldquo;a laboratory experiment to examine whether [password] meters influenced users&amp;rsquo; password selections when they were forced to change their real passwords&amp;rdquo;. The 47 participants didn&amp;rsquo;t know that their passwords were the subject of the study, and the meters yielded significantly stronger passwords.&lt;/p&gt;
&lt;p&gt;A follow-up field experiment with 541 participants tested a different scenario: creating a password for an unimportant account. There the meters made no observable difference, as participants simply reused weak passwords from similar low-risk accounts. The authors conclude that meters help &amp;ldquo;when users are forced to change existing passwords on &amp;lsquo;important&amp;rsquo; accounts&amp;rdquo; and that individual meter design decisions likely have a marginal impact.&lt;/p&gt;
&lt;p&gt;&lt;a href="https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/chi13b.pdf" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="analyzing-password-strength-meters"&gt;Analyzing Password-Strength Meters &lt;a class="heading-anchor" href="#analyzing-password-strength-meters" aria-label="Link to section: Analyzing Password-Strength Meters"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;Researchers from Concordia University tested 13 password meters from 11 popular web services, including Google, Yahoo! and PayPal, with nearly 4 million passwords from password dictionaries. They found highly inconsistent ratings for the same password across meters and many weak passwords labeled as strong or very strong. The study is from 2014, but I still see most of these problems on live sites. The authors argue that meter design deserves real expertise and research, and that, at least, &amp;ldquo;the current meters should avoid providing misleading strength outcomes, especially for weak passwords.&amp;rdquo;&lt;/p&gt;
&lt;p&gt;&lt;a href="https://xavier2dc.fr/papers/password-meters-ndss2014.pdf" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="low-budget-password-strength-estimation"&gt;Low-Budget Password Strength Estimation &lt;a class="heading-anchor" href="#low-budget-password-strength-estimation" aria-label="Link to section: Low-Budget Password Strength Estimation"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;In this reference paper for the zxcvbn strength estimator from Dropbox, Daniel Lowe Wheeler shows that 1.5 MB of compressed storage is enough to accurately estimate the best-known guessing attacks up to 10^5 guesses. With 245 kB and 29 kB, zxcvbn covers 10^4 and 10^3 guesses respectively. The paper compares the estimator against 4 modern guessing attacks and presents it as a better replacement for composition requirements.&lt;/p&gt;
&lt;p&gt;&lt;a href="https://www.usenix.org/system/files/conference/usenixsecurity16/sec16_paper_wheeler.pdf" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="modeling-password-guessability-using-neural-networks"&gt;Modeling Password Guessability Using Neural Networks &lt;a class="heading-anchor" href="#modeling-password-guessability-using-neural-networks" aria-label="Link to section: Modeling Password Guessability Using Neural Networks"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;In 2016, researchers from Carnegie Mellon University showed that artificial neural networks can often guess passwords more effectively than probabilistic context-free grammars and Markov models. The networks can be compressed to as little as hundreds of kilobytes without substantially worsening guessing effectiveness. With this result, the authors built the first principled client-side model of password guessing in JavaScript.&lt;/p&gt;
&lt;blockquote&gt;
&lt;p&gt;&amp;ldquo;While we measured client-side strength metrics based on guessing effectiveness, a remaining challenge is giving user-interpretable advice to improve passwords during password creation.&amp;rdquo;&lt;/p&gt;
&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://www.usenix.org/system/files/conference/usenixsecurity16/sec16_paper_melicher.pdf" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="practical-recommendations-for-stronger-more-usable-passwords"&gt;Practical Recommendations for Stronger, More Usable Passwords &lt;a class="heading-anchor" href="#practical-recommendations-for-stronger-more-usable-passwords" aria-label="Link to section: Practical Recommendations for Stronger, More Usable Passwords"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;In their 2020 study, Carnegie Mellon University researchers ran 2 online experiments and formulated &amp;ldquo;concrete recommendations for policy configurations that produce a good balance of security and usability&amp;rdquo;. They give clear guidance on composition requirements, blocklists, and minimum-strength requirements driven by neural-network strength estimates.&lt;/p&gt;
&lt;blockquote&gt;
&lt;p&gt;&amp;ldquo;We recommend a 1c12+NN10 minimum-strength policy for organizations that wish to protect high-value accounts without a substantial negative usability impact.&amp;rdquo;&lt;/p&gt;
&lt;/blockquote&gt;
&lt;p&gt;In plain terms, the recommended policy requires at least 12 characters and a password that passes the neural-network strength check. I consider this paper one of the most comprehensive and influential collections of best practices for password policies.&lt;/p&gt;
&lt;p&gt;&lt;a href="https://dl.acm.org/doi/pdf/10.1145/3372297.3417882" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h2 id="related"&gt;Related &lt;a class="heading-anchor" href="#related" aria-label="Link to section: Related"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-reuse/"&gt;How common is password reuse?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/credential-stuffing/"&gt;How do you protect accounts from stolen credentials?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/password-manager-friendly/"&gt;Password-manager-friendly login forms&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>Do login challenges and risk scoring stop account takeover?</title><link>https://www.janbrennenstuhl.eu/knowledge-base/account-security/reinforced-authentication/</link><guid isPermaLink="true">https://www.janbrennenstuhl.eu/knowledge-base/account-security/reinforced-authentication/</guid><pubDate>Thu, 27 Apr 2023 00:00:00 +0200</pubDate><description>&lt;p&gt;&lt;em&gt;Latest change (2026-10-13): Rewritten as a field-guide entry: question title, short answer, key findings and evidence table. Study summaries kept and extended with numbers from the papers.&lt;/em&gt;&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; They are an important barrier if you pick the right challenge: device-based challenges blocked over 94% of phishing-rooted hijacking attempts at Google, while knowledge-based challenges stopped as few as 10%. The cost is friction, as 52% of legitimate users who got a challenge failed to sign in at first.&lt;/p&gt;&lt;h2 id="key-findings"&gt;Key findings &lt;a class="heading-anchor" href="#key-findings" aria-label="Link to section: Key findings"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;figure class="finding" id="f-login-device-challenge-94pct"&gt;
&lt;blockquote&gt;&lt;p&gt;Device-based login challenges blocked over 94% of hijacking attempts rooted in phishing and 100% of automated hijacking attempts.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://damonmccoy.com/papers/loginchallenges.pdf" class="ext"&gt;&lt;cite&gt;Doerfler et al., Evaluating Login Challenges as a Defense Against Account Takeover (WWW 2019)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-login-device-challenge-94pct" title="#f-login-device-challenge-94pct"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Device-based login challenges blocked over 94% of hijacking attempts rooted in phishing and 100% of automated hijacking attempts." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/reinforced-authentication/#f-login-device-challenge-94pct" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/reinforced-authentication/f-login-device-challenge-94pct.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-login-knowledge-challenge-10pct"&gt;
&lt;blockquote&gt;&lt;p&gt;Knowledge-based login challenges prevented as few as 10% of hijacking attempts rooted in phishing.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://damonmccoy.com/papers/loginchallenges.pdf" class="ext"&gt;&lt;cite&gt;Doerfler et al., Evaluating Login Challenges as a Defense Against Account Takeover (WWW 2019)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-login-knowledge-challenge-10pct" title="#f-login-knowledge-challenge-10pct"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Knowledge-based login challenges prevented as few as 10% of hijacking attempts rooted in phishing." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/reinforced-authentication/#f-login-knowledge-challenge-10pct" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/reinforced-authentication/f-login-knowledge-challenge-10pct.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-login-challenge-friction-52pct"&gt;
&lt;blockquote&gt;&lt;p&gt;Login challenges caused 52% of legitimate users to fail to sign in, although 97% got into their account within a short period.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://damonmccoy.com/papers/loginchallenges.pdf" class="ext"&gt;&lt;cite&gt;Doerfler et al., Evaluating Login Challenges as a Defense Against Account Takeover (WWW 2019)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-login-challenge-friction-52pct" title="#f-login-challenge-friction-52pct"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="Login challenges caused 52% of legitimate users to fail to sign in, although 97% got into their account within a short period." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/reinforced-authentication/#f-login-challenge-friction-52pct" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/reinforced-authentication/f-login-challenge-friction-52pct.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-login-risk-score-recall-89pct"&gt;
&lt;blockquote&gt;&lt;p&gt;A statistical login classifier using IP and user-agent history reached a recall of up to 89% at a 10% false-positive rate on LinkedIn login data.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.ndss-symposium.org/wp-content/uploads/2017/09/who-are-you-statistical-approach-measuring-user-authenticity.pdf" class="ext"&gt;&lt;cite&gt;Freeman et al., Who Are You? A Statistical Approach to Measuring User Authenticity (NDSS 2016)&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-login-risk-score-recall-89pct" title="#f-login-risk-score-recall-89pct"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="A statistical login classifier using IP and user-agent history reached a recall of up to 89% at a 10% false-positive rate on LinkedIn login data." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/reinforced-authentication/#f-login-risk-score-recall-89pct" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/reinforced-authentication/f-login-risk-score-recall-89pct.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;figure class="finding" id="f-login-protect-frequent-guesses"&gt;
&lt;blockquote&gt;&lt;p&gt;If you can implement only one defence against online guessing, protect users whose passwords attackers guess frequently more aggressively than others.&lt;/p&gt;&lt;/blockquote&gt;
&lt;figcaption&gt;&lt;span class="finding-source"&gt;Source: &lt;a href="https://www.microsoft.com/en-us/research/uploads/prod/2018/06/stopGuessingTR.pdf" class="ext"&gt;&lt;cite&gt;Tian, Herley and Schechter, StopGuessing: Using Guessed Passwords to Thwart Online Guessing&lt;/cite&gt;&lt;/a&gt;&lt;/span&gt;&lt;p class="finding-tools"&gt;&lt;a class="finding-id" href="#f-login-protect-frequent-guesses" title="#f-login-protect-frequent-guesses"&gt;Link&lt;span class="sr-only"&gt; to this finding&lt;/span&gt;&lt;/a&gt;&lt;button type="button" class="text-button" data-quote="If you can implement only one defence against online guessing, protect users whose passwords attackers guess frequently more aggressively than others." data-quote-url="https://www.janbrennenstuhl.eu/knowledge-base/account-security/reinforced-authentication/#f-login-protect-frequent-guesses" hidden&gt;Copy quote&lt;/button&gt;&lt;a class="finding-card" href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/reinforced-authentication/f-login-protect-frequent-guesses.png" download&gt;Save as image&lt;span class="sr-only"&gt; (PNG, 1200 by 630 pixels)&lt;/span&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/figcaption&gt;
&lt;/figure&gt;
&lt;h2 id="evidence"&gt;Evidence &lt;a class="heading-anchor" href="#evidence" aria-label="Link to section: Evidence"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;div class="table-scroll studies" tabindex="0" role="region" aria-label="Reinforced authentication studies"&gt;&lt;table&gt;
&lt;caption&gt;Reinforced authentication studies&lt;/caption&gt;
&lt;thead&gt;&lt;tr&gt;&lt;th scope="col"&gt;Study&lt;/th&gt;&lt;th scope="col"&gt;Year&lt;/th&gt;&lt;th scope="col"&gt;Sample&lt;/th&gt;&lt;th scope="col"&gt;Method&lt;/th&gt;&lt;th scope="col"&gt;Finding&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://damonmccoy.com/papers/loginchallenges.pdf" class="ext"&gt;Evaluating Login Challenges as a Defense Against Account Takeover&lt;/a&gt;&lt;span class="org"&gt;Google, New York University&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2019&lt;/td&gt;&lt;td&gt;Over 350,000 hijacking attempts; 1.2 million legitimate users&lt;/td&gt;&lt;td&gt;Case study of 14 challenge types at Google&lt;/td&gt;&lt;td&gt;Device-based challenges work best; 52% of legitimate users failed at first.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.microsoft.com/en-us/research/uploads/prod/2018/06/stopGuessingTR.pdf" class="ext"&gt;StopGuessing: Using Guessed Passwords to Thwart Online Guessing&lt;/a&gt;&lt;span class="org"&gt;Carnegie Mellon University, Microsoft Research&lt;/span&gt;&lt;/th&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Open-source implementation and attack simulations&lt;/td&gt;&lt;td&gt;Penalize failures with frequently guessed passwords and forgive typos.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.ndss-symposium.org/wp-content/uploads/2017/09/who-are-you-statistical-approach-measuring-user-authenticity.pdf" class="ext"&gt;Who Are You? A Statistical Approach to Measuring User Authenticity&lt;/a&gt;&lt;span class="org"&gt;LinkedIn, Ruhr-Universität Bochum, Università di Cagliari&lt;/span&gt;&lt;/th&gt;&lt;td&gt;2016&lt;/td&gt;&lt;td&gt;LinkedIn login data and simulated attacks&lt;/td&gt;&lt;td&gt;Statistical classifier, prototype evaluation&lt;/td&gt;&lt;td&gt;Recall of up to 89% at a 10% false-positive rate.&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;th scope="row"&gt;&lt;a href="https://www.microsoft.com/en-us/research/uploads/prod/2018/06/distinguishingBotsFromLegit.pdf" class="ext"&gt;Distinguishing Attacks from Legitimate Traffic at an Authentication Server&lt;/a&gt;&lt;span class="org"&gt;Microsoft Research&lt;/span&gt;&lt;/th&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;n/a&lt;/td&gt;&lt;td&gt;Statistical estimation with sensitivity analysis&lt;/td&gt;&lt;td&gt;Estimating the bad-to-good traffic ratio avoids fixed lockout thresholds.&lt;/td&gt;&lt;/tr&gt;
&lt;/tbody&gt;&lt;/table&gt;&lt;/div&gt;
&lt;h2 id="what-to-do"&gt;What to do &lt;a class="heading-anchor" href="#what-to-do" aria-label="Link to section: What to do"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="for-engineers"&gt;For engineers &lt;a class="heading-anchor" href="#for-engineers" aria-label="Link to section: For engineers"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;ul&gt;
&lt;li&gt;Score every login on signals the user doesn&amp;rsquo;t type, such as IP address history, device and browser. Freeman et al. reached a recall of up to 89% using IP and user-agent history.&lt;/li&gt;
&lt;li&gt;Prefer device-based challenges, such as a prompt on a known device, over knowledge questions. Doerfler et al. measured a large gap between the two. My essay on &lt;a href="https://www.janbrennenstuhl.eu/step-up-authentication/"&gt;step-up authentication&lt;/a&gt; shows how to request a stronger factor only when the risk justifies it.&lt;/li&gt;
&lt;li&gt;Count failed logins by how likely they are to be an attack. Penalize failures with frequently guessed passwords, protect accounts with weak passwords harder, and be lenient with typos. Tian et al. do this without storing anything that would expose accounts if leaked.&lt;/li&gt;
&lt;li&gt;Breadth-first guessing spreads attempts thinly over many accounts, so per-account lockouts don&amp;rsquo;t see it. Herley and Schechter estimate the share of attack traffic instead of relying on fixed thresholds. See &lt;a href="https://www.janbrennenstuhl.eu/password-spraying/"&gt;password spraying&lt;/a&gt; and &lt;a href="https://www.janbrennenstuhl.eu/honeytoken/"&gt;honeytokens&lt;/a&gt; for related detection.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="for-product-and-ux"&gt;For product and UX &lt;a class="heading-anchor" href="#for-product-and-ux" aria-label="Link to section: For product and UX"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;ul&gt;
&lt;li&gt;Budget for friction. A challenge that half of your legitimate users fail at first needs a clear path back in, or it turns into support tickets and lost users.&lt;/li&gt;
&lt;li&gt;Challenge only suspicious logins, so most users never see a challenge. The 2 risk-scoring studies above show how to pick them out.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="the-studies-in-detail"&gt;The studies in detail &lt;a class="heading-anchor" href="#the-studies-in-detail" aria-label="Link to section: The studies in detail"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;h3 id="evaluating-login-challenges-as-a-defense-against-account-takeover"&gt;Evaluating Login Challenges as a Defense Against Account Takeover &lt;a class="heading-anchor" href="#evaluating-login-challenges-as-a-defense-against-account-takeover" aria-label="Link to section: Evaluating Login Challenges as a Defense Against Account Takeover"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;Google and New York University teamed up to &amp;ldquo;study the efficacy of login challenges at preventing account takeover&amp;rdquo; and to &amp;ldquo;evaluate the amount of friction these challenges create for normal users&amp;rdquo;. Using Google as a case study, they evaluated 14 device-based, delegation-based, knowledge-based and resource-based challenges against over 350,000 real-world hijacking attempts, and measured usability on a sample of 1.2 million legitimate users.&lt;/p&gt;
&lt;p&gt;Device-based challenges blocked over 94% of phishing-rooted attempts and all automated ones. Knowledge-based challenges stopped as few as 10% of phishing-rooted attempts and 73% of automated ones. The conclusion is sobering:&lt;/p&gt;
&lt;blockquote&gt;
&lt;p&gt;&amp;ldquo;Our results illustrate that login challenges act as an important barrier to hijacking, but that friction in the process leads to 52% of legitimate users failing to sign-in—though 97% of users eventually access their account in a short period.&amp;rdquo;&lt;/p&gt;
&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://damonmccoy.com/papers/loginchallenges.pdf" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="using-guessed-passwords-to-thwart-online-guessing"&gt;Using Guessed Passwords to Thwart Online Guessing &lt;a class="heading-anchor" href="#using-guessed-passwords-to-thwart-online-guessing" aria-label="Link to section: Using Guessed Passwords to Thwart Online Guessing"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;This paper from Microsoft Research and Carnegie Mellon University gives practitioners who defend password-protected resources against online guessing a set of techniques to better balance friction and security:&lt;/p&gt;
&lt;blockquote&gt;
&lt;p&gt;&amp;ldquo;Among our contributions are features to make better benign/attack traffic classification decisions. These include mechanisms to penalize fail events with commonly guessed passwords more, to protect accounts with weak passwords better, and to distinguish user-generated typos from other fails. We do all of these without storing any information which exposes accounts to new risk in the event of a server breach. These innovations allow much greater flexibility over the previous state-of-the-art, which treats all failures equally&amp;rdquo;&lt;/p&gt;
&lt;/blockquote&gt;
&lt;p&gt;The authors publish an open-source implementation and a guessing-attack simulator. If practitioners can implement only one of the techniques, they recommend protecting users who rely on frequently guessed passwords more aggressively.&lt;/p&gt;
&lt;p&gt;&lt;a href="https://www.microsoft.com/en-us/research/uploads/prod/2018/06/stopGuessingTR.pdf" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="a-statistical-approach-to-measuring-user-authenticity"&gt;A Statistical Approach to Measuring User Authenticity &lt;a class="heading-anchor" href="#a-statistical-approach-to-measuring-user-authenticity" aria-label="Link to section: A Statistical Approach to Measuring User Authenticity"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;Researchers from LinkedIn, Ruhr-Universität Bochum and Università di Cagliari evaluated a statistical framework for &amp;ldquo;strengthening password-based authentication by classifying login attempts into normal and suspicious activity based on parameters available during login&amp;rdquo;. Suspicious attempts can then get an extra verification step, such as a phone-based one.&lt;/p&gt;
&lt;p&gt;They tested the approach on real-life login data from LinkedIn and on simulated attacks. Using the user&amp;rsquo;s IP history and user-agent history, it showed a &amp;ldquo;recall of up to 89% for a false-positive rate of 10%&amp;rdquo;.&lt;/p&gt;
&lt;p&gt;&lt;a href="https://www.ndss-symposium.org/wp-content/uploads/2017/09/who-are-you-statistical-approach-measuring-user-authenticity.pdf" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="distinguishing-attacks-from-legitimate-traffic-at-an-authentication-server"&gt;Distinguishing Attacks from Legitimate Traffic at an Authentication Server &lt;a class="heading-anchor" href="#distinguishing-attacks-from-legitimate-traffic-at-an-authentication-server" aria-label="Link to section: Distinguishing Attacks from Legitimate Traffic at an Authentication Server"&gt;#&lt;/a&gt;&lt;/h3&gt;
&lt;p&gt;In this paper, Microsoft researchers look at how to defend against breadth-first attacks that spread guesses very widely. Lockout rules such as &amp;ldquo;three strikes&amp;rdquo; work against attacks on one account, but at large providers breadth-first attacks can send millions of guesses without triggering them. The authors describe &amp;ldquo;a simple robust procedure to estimate the ratio of bad-to-good traffic and show how this can be used to calculate the likelihood that any particular observation is indicative of malice.&amp;rdquo;&lt;/p&gt;
&lt;p&gt;They conclude that their &amp;ldquo;approach has the advantage over three strikes type lockout and variants that it has no dependence on arbitrary thresholds and we avoid base rate neglect. It has the advantages over machine learning schemes that it requires no labels, and does not assume stationarity of attack traffic.&amp;rdquo;&lt;/p&gt;
&lt;p&gt;&lt;a href="https://www.microsoft.com/en-us/research/uploads/prod/2018/06/distinguishingBotsFromLegit.pdf" class="ext"&gt;Read the paper (PDF)&lt;/a&gt;&lt;/p&gt;
&lt;h2 id="related"&gt;Related &lt;a class="heading-anchor" href="#related" aria-label="Link to section: Related"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/credential-stuffing/"&gt;How do you protect accounts from stolen credentials?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/account-takeover/"&gt;How do attackers take over accounts?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.janbrennenstuhl.eu/knowledge-base/account-security/password-reuse/"&gt;How common is password reuse?&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</description></item></channel></rss>