karakeep CVE-2026-75773

The login endpoint is network-reachable and requires no prior authentication or session state, making the attack surface accessible to any remote actor (PR:N in both CVSS assessments). No special configuration, enabled feature flag, or non-default deployment mode is required to reach the vulnerable endpoint. However, successful exploitation - meaning actual account compromise - requires that the targeted account holds a password guessable within a brute-force or credential-stuffing campaign; this dependency on external password weakness is the basis for the AC:H rating. The attack is further constrained by the need to identify valid email addresses registered on the specific instance. Post-upgrade, exploitation is limited by the additional condition that the server's rateLimiting.enabled flag must be false (default or misconfigured) for the vulnerability to persist in version 0.33.0.

The CVSS 4.0 base vector AV:N/AC:H/AT:N/PR:N/UI:N/VC:L/VI:N/VA:N yields a score of 2.9, placing this at the low end of severity. The AC:H rating is consistent with the VulDB characterization of 'difficult' exploitability: while the login endpoint is unauthenticated and network-reachable, successfully compromising an account requires guessing a valid password - a variable the attacker does not control. No EPSS data was supplied for independent probability comparison, and the CVE does not appear in CISA KEV, confirming no confirmed opportunistic or widespread exploitation. The public POC (GitHub issue #2919) modestly elevates real-world risk by reducing attacker effort for initial tooling. Impact is bounded to limited confidentiality (VC:L): a successful attacker gains access only to the targeted user's bookmark data, with no integrity or availability consequences reported. A critical operational nuance: even organizations that upgrade to 0.33.0 remain vulnerable unless rateLimiting.enabled is explicitly enabled in the server configuration - this conditional fix behavior is not typical and warrants verification during remediation.

An attacker enumerates a publicly exposed karakeep instance and identifies a registered user's email address from a prior data breach or public profile. Using an automated credential-stuffing tool seeded with commonly leaked passwords, the attacker submits thousands of unauthenticated POST requests to the login endpoint with no server-side throttle to interrupt them. If any target account uses a weak or previously breached password, the tool obtains a valid session, granting full access to that user's saved bookmarks and read-later content. A publicly available proof-of-concept (GitHub issue #2919) provides a baseline for tooling this attack.