Repository navigation
Hall of Fame
Lauri Ojansivu edited this page Nov 8, 2026
·
4 revisions
Security issues found in, and fixed in, caddy-wekan-plugin - WeKan written in Go inside Caddy, part of Weddy. The same columns as WeKan's Hall of Fame. Report a vulnerability privately as described in WeKan's SECURITY.md; stars follow WeKan's star audit. Findings of automatic scanners have no stars.
| CVE | Icon | Vulnerability name | Date | Responsible Security Disclosure by | Stars | Process | Vulnerabilities |
|---|---|---|---|---|---|---|---|
| CVE-2026-116587, CVE-2026-116585, CVE-2026-116586 | 🔑📦🩸 | JoseBleed | 2026-11-08 | GitHub Dependabot and xet7 | - | ProcessDependabot alerts #1, #2 and #3 for github.com/go-jose/go-jose/v3 <= 3.0.5, which has no patched release. It came in through Caddy's caddytls, smallstep/certificates and go.step.sm/crypto; go.step.sm/crypto v0.92.0 and smallstep/certificates at its main branch (the commit Caddy's master uses) moved to go-jose v4.1.9, and the plugin now requires them, so go-jose v3 is no longer in the build. |
DetailsThree denial-of-service issues in go-jose v3: a crafted RSA key in a JWK costs excess CPU (GHSA-mgcc-cww5-fp2q), CBC-HMAC decryption can panic (GHSA-4w3g-jfj2-cw2q), and deeply nested JSON can exhaust the stack (GHSA-vvm8-gf78-28g4). They are in Caddy's ACME/step-CA code, not in WeKan's. Fix: commit. |
| - | 🔎🗄️🩸 | QueryBleed | 2026-11-08 | GitHub CodeQL and xet7 | - | ProcessCode scanning alerts #1-#5 (go/sql-injection, "Database query built from user-controlled sources") at web/server.go and web/auth.go; a source review found the same shape in the REST API, the card's list and swimlane lookup and the next sort position, and fixed them too. |
DetailsIds from a request's address, form or JSON body - board, card, list, swimlane, user - and the sign-in name went into MongoDB filters as they came. In Go they are always strings, so no operator could be injected, but nothing said so. Now an id is used only when it looks like one (letters, digits, _ and -, at most 64), every request value is compared with$eq, and a sign-in name is at most 254 characters; anything else is not found. A test reads every file of the package and fails on a filter value that is neither eq(...) nor a database id. Fix: commit. |
| - | 🔐🔢🩸 | Argon2WidthBleed | 2026-11-08 | GitHub CodeQL and xet7 | - | ProcessCode scanning alerts #6 and #7 (go/incorrect-integer-conversion) at web/auth.go. |
DetailsThe parameters of a stored argon2 hash were parsed as 32-bit numbers and then converted to argon2's widths: a parallelism above 255 was cut short (256 became 0), and memory and passes had no upper bound, so a hash in the database could make a sign-in take any amount of memory and time. Each parameter is now parsed at argon2's own width and bounded (1 GiB, 16 passes, parallelism 1-255, a 16-128 byte hash) before anything is computed. Fix: commit. |
| - | 🍪🔓🩸 | CookieBleed | 2026-11-08 | GitHub CodeQL and xet7 | - | ProcessCode scanning alert #8 (go/cookie-secure-not-set) at web/server.go, the cookie that clears the login; the login cookie itself was Secure only when the request came in over TLS. |
DetailsThe meteor_login_token cookie, which carries the login token, is now always Secure, HttpOnly and SameSite=Lax, and so is the one logging out sets - the token never travels over plain HTTP. Caddy gives every site HTTPS, and browsers keep Secure cookies on http://localhost and http://127.0.0.1, so local tests still work. Fix: commit. |
| - | #️⃣🛡️ | PreHash (not a vulnerability) | 2026-11-08 | GitHub CodeQL and xet7 | - | ProcessCode scanning alert #9 (go/weak-sensitive-data-hashing) at web/auth.go, closed the way WeKan closed its #599 and FerretDB its SCRAM-SHA-1 alert: by the code. |
DetailsSHA-256 of the password is Meteor accounts-password's pre-hash, the input bcrypt (cost 10) or argon2 is given - not how the password is stored; removing it would make every existing WeKan password stop matching. It is now meteorPreHash, in a file of its own, saying what it is and is not, computed with crypto.SHA256.New() from the crypto package's hash registry. TestNoFastHashOfPasswords pins that it is Meteor's digest, that crypto.SHA256 is used nowhere else and that no other fast hash touches a password; TestPasswordsAreStoredSlowlyHashed that what is stored is bcrypt of cost 10 or more. Fix: commit. |