CVE-2026-33540

hi guys,

commit: 40594bd98e6d6ed993b5c6021c93fdf96d2e5851 (as-of 2026-01-31)
contact: GitHub Security Advisory (https://github.com/distribution/distribution/security/advisories/new)

in pull-through cache mode, distribution discovers token auth endpoints by parsing WWW-Authenticate challenges returned by the configured upstream registry. the realm URL from a bearer challenge is used without validating that it matches the upstream registry host. as a result, an attacker-controlled upstream (or an attacker with MitM position to the upstream) can cause distribution to send the configured upstream credentials via basic auth to an attacker-controlled realm URL.

this is the same vulnerability class as CVE-2020-15157 (containerd), but in distribution’s pull-through cache proxy auth flow.

severity

HIGH

note: the baseline impact is credential disclosure of the configured upstream credentials. if a deployment uses broader credentials for upstream auth (for example cloud iam credentials), the downstream impact can be higher; i am not claiming this as default for all deployments.

affected components

• registry/proxy/proxyauth.go:66-81 (getAuthURLs): extracts bearer realm from upstream WWW-Authenticate without validating destination
• internal/client/auth/session.go:485-510 (fetchToken): uses the realm URL directly for token fetch
• internal/client/auth/session.go:429-434 (fetchTokenWithBasicAuth): sends credentials via basic auth to the realm URL

reproduction

attachment: poc.zip (local harness) with canonical and control runs.

the harness is local and does not contact a real registry: it uses two local HTTP servers (upstream + attacker token service) to demonstrate whether basic auth is sent to an attacker-chosen realm.

unzip -q -o poc.zip -d poc
cd poc
make canonical
make control

expected output (excerpt):

[CALLSITE_HIT]: getAuthURLs::configureAuth
[PROOF_MARKER]: basic_auth_sent=true realm_host=127.0.0.1 account_param=user authorization_prefix=Basic

control output (excerpt):

[CALLSITE_HIT]: getAuthURLs::configureAuth
[NC_MARKER]: realm_validation=PASS basic_auth_sent=false

suggested remediation

validate that the token realm destination is within the intended trust boundary before associating credentials with it or sending any authentication to it. one conservative option is strict same-host binding: only accept a realm whose host matches the configured upstream host.

fix accepted when

• distribution does not send configured upstream credentials to an attacker-chosen realm URL
• a regression test covers the canonical and blocked cases

addendum.md
poc.zip
PR_DESCRIPTION.md
RUNNABLE_POC.md

best,
oleh

credential exfiltration of the upstream authentication material configured for the pull-through cache.

attacker starting positions that make this realistic:

• supply chain / configuration: an operator configures a proxy cache to use an upstream that becomes attacker-controlled (compromised registry, stale domain, or a malicious mirror)
• network: MitM on the upstream connection in environments where the upstream is reachable over insecure transport or a compromised network path

Untrusted input controls the target URL of a server-initiated request, which may reach internal services not otherwise accessible from outside. Typical impact: access to internal metadata services, internal APIs, or cloud credentials.

CVE-2026-33540 has a CVSS score of 7.5 (High). The vector is network-reachable, no privileges required, and no user interaction. A CVSS score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether this affects your application depends on whether the vulnerable code is present and reachable in your environment. A fixed version is available (3.1.0); upgrading removes the vulnerable code path.

Kodem intelligence

1. What is CVE-2026-33540? CVE-2026-33540 is a high-severity server-side request forgery (SSRF) vulnerability in github.com/distribution/distribution/v3 (go), affecting versions < 3.1.0. It is fixed in 3.1.0. Untrusted input controls the target URL of a server-initiated request, which may reach internal services not otherwise accessible from outside.
2. How severe is CVE-2026-33540? CVE-2026-33540 has a CVSS score of 7.5 (High). This score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether it represents real risk in your environment depends on whether the vulnerable code is present and reachable.
3. Which packages are affected by CVE-2026-33540?
   • github.com/distribution/distribution/v3 (go) (versions < 3.1.0)
   • github.com/distribution/distribution (go) (versions <= 2.8.3)
4. Is there a fix for CVE-2026-33540? Yes. CVE-2026-33540 is fixed in 3.1.0. Upgrade to this version or later.
5. Is CVE-2026-33540 exploitable, and should I be worried? Whether CVE-2026-33540 is exploitable in your environment depends on whether the vulnerable code is present and reachable. A CVSS score is a worst-case rating; it does not account for your specific deployment, configuration, or usage patterns. Kodem, an Intelligent Application Security platform, uses runtime intelligence to show which vulnerabilities actually execute in production, so you can focus on the ones that represent real risk. Get a demo
6. What actually determines whether CVE-2026-33540 is exploitable, and how bad it is? Exploitability and impact are not fixed properties of a CVE. They depend on runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A high CVSS score on a dependency that never runs is not the same as real risk. Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter.
7. How do I fix CVE-2026-33540? Upgrade github.com/distribution/distribution/v3 to 3.1.0 or later.