n8n CVE-2026-59208

Exploitation requires the n8n instance to be configured with more than one trusted token-exchange issuer - a single-issuer or non-federated deployment is not vulnerable. The attacker must hold a valid token from one of the trusted issuers (PR:L, authenticated per CVSS PR:L) and must possess or provision a sub claim that exactly matches a victim account registered under a different trusted issuer (AT:P - a cross-issuer sub collision must exist and be attacker-achievable). Limiting factors: the non-default multi-issuer configuration, the need for a valid trusted-issuer token, and the requirement to know or control a colliding sub value all constrain who can actually carry out the attack; no user interaction from the victim is needed.

The supplied CVSS 4.0 vector (AV:N/AC:L/AT:P/PR:L/UI:N/VC:H/VI:H/VA:N, base 7.6 High) tells a consistent story: network-reachable, no user interaction, but requiring some privilege (PR:L - the attacker must already hold a valid token from one trusted issuer) and an attack requirement (AT:P - the target environment must run multiple trusted issuers and a sub collision must exist). Confidentiality and integrity of the victim account are High while availability is unaffected (VA:N), matching an account-takeover rather than a service-disruption bug. No EPSS score, KEV listing, POC, or SSVC decision data was provided; treat exploitation likelihood as unquantified from this input. This is a real priority for organizations running n8n with federated multi-issuer SSO, but the AT:P/PR:L preconditions meaningfully narrow the exposed population - default single-issuer or non-SSO instances are not affected, so it is not a mass-exploitable internet-wide issue.

An attacker who legitimately controls (or can register) an account at trusted Issuer B sets or obtains a sub value identical to a high-privilege victim's sub under trusted Issuer A. They present a valid token-exchange token from Issuer B to the multi-issuer n8n instance; n8n matches on sub alone, ignores the iss, and logs the attacker into the victim's account with the victim's workflows and credentials. No user interaction is required and no public exploit is identified at time of analysis.