Patch Forecast
Yellow warning
Plan into this sprint.

GHSA-4hv6-xc92-j86g

Vikunja: WebSocket authentication ignores server-side session state, so revoked sessions keep receiving live pushes

Risk score
44/100
CVSS
6.5 v3.1
Published
9 Oct 2026

Open in the live forecast Today's forecast

What to know

What it is
Vikunja v2.6.0 introduced server-side sessions with revocation semantics: DELETE /api/v2/user/sessions/{id} documents "Revokes a specific session by its UUID", and enabling TOTP calls DeleteAllUserSessions to invalidate all sessions. The WebSocket endpoint accepts any cryptographically valid user JWT without ever reso…
Affected
code.vikunja.io/api (go) >= 2.3.0, <= 2.6.0
Fix
No patched version listed yet: check the advisory for mitigations.
Exploitation
No known exploitation

Story map

Story map: code.vikunja.io/apiTHE FLAWSTHE FIXGHSA-4hv6-xc92-j86gVikunja: WebSocketauthentication ignoresserver-side session state, so…CVSS 6.5CVE-2026-91972Vikunja: Every /api/v2pre-auth endpoint isunthrottled on a stock install…CVSS 8.7 · EPSS 1%CVE-2026-68581Vikunja: Link-share principalID collision allowscross-account API token…CVSS 8.1 · EPSS 0%CVE-2026-62376Vikunja: Plaintext storage ofpassword-reset/email-confirmtokens in database enables…CVSS 8.1 · EPSS 0%Upgrade code.vikunja.io/api2.4.0or later on your release lineWho is exposedcode.vikunja.io/api (go) <= 2.5.0AMBER · PATCH THIS WEEKcode.vikunja.io/apiregisterAPIRoutesV2 neverapplies the unconditionalpre-auth rate-limit floor(unauthRateLimit()) to the v2public routes — it passes thatlimiter only to /api/v2/ws — andotherwise relies on…Where it standsPublic proof of conceptCONNECTS TOSame weaknessCWE-613 (CWE-613)34 other warnings in 30 days →
Story map: the flaws, the fix, how it unfolded and what it connects to. Boxes link to the record or article.

At a glance

code.vikunja.io/api: Amber, patch this week. registerAPIRoutesV2 never applies the unconditional pre-auth rate-limit floor (unauthRateLimit()) to the v2 public routes — it passes that limiter only to /api/v2/ws — and otherwise relies on setupRateLimit, which registers nothing when ratelimit.enabled is false (the default). So on a stock install every v2 pre-auth …

The flaws

  • GHSA-4hv6-xc92-j86g: Vikunja: WebSocket authentication ignores server-side session state, so revoked sessions keep receiving live pushes (CVSS 6.5)
  • CVE-2026-91972: Vikunja: Every /api/v2 pre-auth endpoint is unthrottled on a stock install while its /api/v1 twin is rate limited (CVSS 8.7 · EPSS 1%)
  • CVE-2026-68581: Vikunja: Link-share principal ID collision allows cross-account API token issuance and management (CVSS 8.1 · EPSS 0%)
  • CVE-2026-62376: Vikunja: Plaintext storage of password-reset/email-confirm tokens in database enables account takeover on DB read access (CVSS 8.1 · EPSS 0%)

The fix

Upgrade code.vikunja.io/api: 2.4.0 or later on your release line

Who is exposed: code.vikunja.io/api (go) <= 2.5.0

Where it stands

  • Public proof of concept

Connects to

Why it is Yellow

  • CVSS 6.5 medium
  • On your watchlist: Go

Affected packages

PackageEcosystemVulnerableFixed in
code.vikunja.io/apigo>= 2.3.0, <= 2.6.0No fix yet

References

Original description

### Summary Vikunja v2.6.0 introduced server-side sessions with revocation semantics: `DELETE /api/v2/user/sessions/{id}` documents "Revokes a specific session by its UUID", and enabling TOTP calls `DeleteAllUserSessions` to invalidate all sessions. The WebSocket endpoint accepts any cryptographically valid user JWT without ever resolving its `sid` (session id) claim against the sessions table. A connection authenticated before revocation stays authenticated and continues to receive live pushes indefinitely, and even brand-new WebSocket connections are accepted with a token whose session was deleted. The revocation feature therefore does not cover the WebSocket boundary at all. ### Details User access JWTs carry a session reference: `NewUserJWTAuthtoken` (`pkg/modules/auth/auth.go:188`) sets `claims["sid"] = sessionID`, and the API login-session documentation describes the claim as tying the JWT to a server-side session. On the HTTP side nothing consumes the claim per request either (REST access tokens simply expire after their short TTL, a design the maintainers documented when fixing GHSA-96q5, and refresh tokens are correctly invalidated when the session row is deleted). The WebSocket endpoint `GET /api/v2/ws` authenticates through the first client message. `Connection.handleAuth` (`pkg/websocket/connection.go:168`) validates the token exclusively with `auth.GetUserIDFromToken` (`pkg/modules/auth/auth.go:292`): ```go func GetUserIDFromToken(tokenString string) (int64, error) { token, err := jwt.Parse(tokenString, func(_ *jwt.Token) (any, error) { return []byte(config.ServiceSecret.GetString()), nil }) ... typ, ok := claims["type"].(float64) if !ok || int(typ) != AuthTypeUser { return 0, jwt.ErrTokenInvalidClaims } ... } ``` The function verifies signature and token type only. No session lookup occurs anywhere in the WebSocket path, and no re-validation happens after the initial auth. Consequently: 1. A connection authenticated before `DELETE /api/v2/user/sessions/{id}` (or before TOTP enable, which runs `models.DeleteAllUserSessions`, `pkg/routes/api/v2/user_totp.go:208`) remains registered in the hub and keeps receiving every subscribed event, such as `notification.created` with the full notification object including comment, task and project content. 2. A new connection opened after the deletion, using the same pre-revocation token, authenticates successfully. 3. Session deletion is observable only in the REST surface. The user's session list empties, the refresh token dies on next use, but the WebSocket channel never notices. For REST the exposure is bounded by the 10-minute access-token TTL. For WebSocket it is not: a connection lives until the client closes it, so a session the user believes revoked retains a live, server-pushed data channel with no time bound. ### PoC > [vikunja-ws-auth-ignores-session-revocation-PoC.zip](https://github.com/user-attachments/files/32741065/vikunja-ws-auth-ignores-session-revocation-PoC.zip) The attached archive contains `deployment/deploy.sh` (local Vikunja v2.6.0 on loopback with SQLite) and `poc/poc.sh` running `poc/poc.py`: 1. Victim registers and logs in. Second user registers for later interaction. 2. Control: the victim's WebSocket authenticates and subscribes to `notification.created`. Negative control: a garbage token is rejected with `invalid_token`. 3. Every victim session is deleted via `DELETE /api/v2/user/sessions/{id}` (the list afterwards reports 0 remaining). 4. Exploit A: a brand-new WebSocket connection authenticates successfully with the deleted-session token. 5. Exploit B: the second user comments on a task mentioning the victim. The victim's pre-revocation WebSocket connection receives the `notification.created` push with the full notification payload. Result: 10 checks passed on two consecutive runs, each from a wiped database. ### Impact Session revocation, the mechanism a user relies on when a device is lost or a token is suspected stolen, silently excludes the WebSocket channel. A stolen or leaked access token can be converted into a push channel that survives every revocation action available to the user, including TOTP enrollment whose documented purpose is invalidating all sessions, and keeps delivering the victim's notifications (comment mentions, task and project metadata) until the attacker chooses to disconnect. The fix is to resolve the `sid` claim against the sessions table during `handleAuth` and to re-validate or drop hub connections when their session is deleted.

Last activity 9 Oct 2026. Patch Forecast updates every hour.