CVE-2025-46722
vLLM is an inference and serving engine for large language models (LLMs). In versions starting from 0.7.0 to before 0.9.0, in the file vllm/multimodal/hasher.py, the MultiModalHasher class has a security and data integrity issue in its image hashing method. Currently, it serializes PIL.Image.Image objects using only obj.tobytes(), which returns only the raw pixel data, without including metadata such as the image’s shape (width, height, mode). As a result, two images of different sizes (e.g., 30x100 and 100x30) with the same pixel byte sequence could generate the same hash value. This may lead to hash collisions, incorrect cache hits, and even data leakage or security risks. This issue has been patched in version 0.9.0.
Scoring
- Severity
- MEDIUM
- CVSS base score
- 4.2
- CVSS vector
- CVSS:3.1/AV:N/AC:H/PR:L/UI:N/S:U/C:L/I:N/A:L
- EPSS probability
- 0.32%
- CWE
- CWE-1288, CWE-1023
- Published
- 2025-05-29
- Last modified
- 2026-03-13
Affected products
- vllm-project vllm
Weakness type
Related vulnerabilities
- CVE-2026-69793 — Windows TCP/IP Security Feature Bypass Vulnerability
- CVE-2026-18238 — OOBR in rpcap client in libpcap before 1.10.7
- CVE-2026-18794 — OpenRGB: insufficient input data checks lead to Denial-of-Service, memory overread and overwrite
- CVE-2026-73219 — CVAT: Denial of service with regards to automatic annotation
- CVE-2026-18209 — Keycloak-services: keycloak-services: oidc redirect_uri fragment bypass in http parameter pollution check
- CVE-2026-15943 — Keycloak-services: keycloak-services: oidc idp update reuses masked client secret after token url change
- CVE-2026-42982 — Windows Secure Kernel Mode Elevation of Privilege Vulnerability
- CVE-2026-14781 — Keycloak-services: keycloak-services: oidc email_verified claim incorrectly applied to userinfo email