Impact
When a JWE uses a password-based key-encryption algorithm (PBES2-HS256+A128KW, PBES2-HS384+A192KW, PBES2-HS512+A256KW), PBES2AESKW::unwrapKey() reads the p2c (PBKDF2 iteration count) parameter directly from the attacker-controlled JOSE header and passes it to hashpbkdf2() with no upper bound. The only validation performed (checkHeaderAdditionalParameters()) was isint($p2c) && $p2c > 0.
An unauthenticated attacker can craft a single JWE whose protected header sets a very large p2c (e.g. 100000000 ≈ 87 s of CPU, or PHPINTMAX), forcing a worker to spend an arbitrary amount of CPU inside PBKDF2 before the key unwrap can even fail. The decrypter swallows the eventual exception, so the attacker pays almost nothing while the server burns CPU. JSON General serialization (multiple recipients) and multi-key JWKSets multiply the cost. This is a classic uncontrolled-resource-consumption (CWE-400) denial of service.
Affected configurations
Applications that register any PBES2-HS+AKW algorithm in their decryption AlgorithmManager.
Patches
PBES2AESKW now enforces a configurable maximum iteration count (DEFAULTMAXCOUNT = 1000000, well above realistic legitimate values which are a few thousand) in checkHeaderAdditionalParameters(), before any PBKDF2 computation. The bound is exposed as a constructor argument so operators can tune it.
Workarounds
Before upgrading: validate/limit the p2c header with a custom header checker, or do not enable PBES2 algorithms for untrusted tokens.
References
- RFC 7518 §4.8 (PBES2) - CWE-400: Uncontrolled Resource Consumption
Résolution
Un correctif a été préparé sur une branche dédiée basée sur 3.4.x, avec des tests anti-régression dédiés (fork privé temporaire de cette advisory, PR #1).
PBES2 — PBES2AESKW::unwrapKey() borne désormais le paramètre p2c (constante DEFAULTMAXCOUNT = 1000000, configurable via le constructeur) avant tout appel à hashpbkdf2(), empêchant l'amplification CPU (DoS).
Validation : php -l OK, PHPUnit vert, aucune nouvelle erreur PHPStan introduite (différentiel nul vs 3.4.x), aucun commentaire ajouté dans le code source. Après merge, cascade prévue 3.4.x → 4.0.x → 4.1.x.
web-token jwt-framework and jwt-library before 3.4.12, 4.0.x before 4.0.8, 4.1.x before 4.1.10, and 4.2.x before 4.2.4, and jwt-encryption-algorithm-ecdh-es through 3.2.10, have a ConcatKDF::generate() that performs a single SHA-256 round, deriving at most 32 bytes for ECDH-ES and ECDH-SS direct key agreement. Anyone holding a JWE using A256CBC-HS512 can decrypt it under an all-zero AES key, while A192CBC-HS384 content keys have only 64-bit strength.
Impact
RSACrypt::decryptWithRSA15() (used by the RSA15 key-encryption algorithm) implements RSAES-PKCS1-v15 decryption by inspecting the padding after RSADP and throwing InvalidArgumentException as soon as the padding is malformed. It does not implement the implicit-rejection countermeasure required by RFC 3447 §7.2.2 / RFC 8017 §7.1.2 (return a deterministic pseudo-random value of the expected length on padding failure and let the downstream step fail uniformly).
From a JWE caller this yields a Bleichenbacher/Marvin padding oracle: an attacker submitting adaptively crafted encryptedkey values can distinguish (a) padding rejected, (b) padding valid but wrong CEK length, and (c) padding valid and full AEAD executed — even though JWEDecrypter returns the same false in all cases — because each path performs a measurably different amount of work, amplifiable by enlarging the ciphertext (CWE-208 timing side channel). Enough adaptive queries can recover the wrapped CEK.
Affected configurations
Applications that register RSA15 in their decryption AlgorithmManager and hold an RSA private key.
Patches
PKCS#1 v1.5 decryption now performs implicit rejection: on invalid padding (or unexpected recovered length) it returns a random CEK of the expected size selected in constant time, so the full content-decryption (AEAD) step always runs and fails uniformly, removing the observable difference between padding-valid and padding-invalid ciphertexts. Users are still strongly encouraged to migrate to RSA-OAEP.
Workarounds
Prefer RSA-OAEP/RSA-OAEP-256; do not enable RSA15 for untrusted tokens.
References
- RFC 3447 §7.2.2 / RFC 8017 §7.1.2 - Bleichenbacher (1998); Marvin attack - CWE-208: Observable Timing Discrepancy
Résolution
Un correctif a été préparé sur une branche dédiée basée sur 3.4.x, avec des tests anti-régression dédiés (fork privé temporaire de cette advisory, PR #1).
RSA15 — déchiffrement PKCS#1 v1.5 avec implicit rejection en temps constant (RFC 3447 §7.2.2) : un padding invalide n'est plus distinguable d'un padding valide, neutralisant l'oracle Bleichenbacher.
Validation : php -l OK, PHPUnit vert, aucune nouvelle erreur PHPStan introduite (différentiel nul vs 3.4.x), aucun commentaire ajouté dans le code source. Après merge, cascade prévue 3.4.x → 4.0.x → 4.1.x.
Impact
The experimental Chacha20Poly1305 key-encryption algorithm generates the 16-byte Poly1305 authentication tag during encryptKey() but discards it: the tag is never written to the header and therefore never reaches the wire. On the receiving side, decryptKey() calls openssldecrypt('chacha20-poly1305', ...) without the tag argument, which makes OpenSSL skip authentication entirely.
As a result the AEAD construction is silently degraded to unauthenticated ChaCha20: a tampered encrypted CEK is accepted, and because ChaCha20 is a stream cipher, a single-byte change in the ciphertext propagates as a single-byte change in the recovered CEK with no integrity check (CWE-353 / CWE-347). An attacker on the token path can manipulate the wrapped key without detection.
Affected configurations
Applications that register Jose\Experimental\KeyEncryption\Chacha20Poly1305 (package web-token/jwt-experimental) as a JWE alg.
Patches
encryptKey() now publishes the Poly1305 tag as the base64url tag header parameter (and asserts it is 16 bytes). decryptKey() requires the tag header, validates its length, and passes it to openssldecrypt() so the tag is actually verified, in line with RFC 7539 / RFC 8439. Tampering now results in a decryption failure.
Note: this changes the wire format of tokens produced with this experimental algorithm (a tag header is now emitted and required).
Workarounds
Do not use the experimental Chacha20Poly1305 key-encryption algorithm for untrusted input until upgraded.
References
- RFC 7539 / RFC 8439 (ChaCha20-Poly1305 AEAD) - CWE-353: Missing Support for Integrity Check
Résolution
Un correctif a été préparé sur une branche dédiée basée sur 3.4.x, avec des tests anti-régression dédiés (fork privé temporaire de cette advisory, PR #1).
ChaCha20-Poly1305 — le tag d'authentification Poly1305 est désormais publié dans le header au chiffrement et vérifié au déchiffrement (RFC 7539), rétablissant l'intégrité AEAD.
Validation : php -l OK, PHPUnit vert, aucune nouvelle erreur PHPStan introduite (différentiel nul vs 3.4.x), aucun commentaire ajouté dans le code source. Après merge, cascade prévue 3.4.x → 4.0.x → 4.1.x.