vLLM up to 0.23.1rc0 API Image error.message log file

Summaryinfo

A vulnerability was found in vLLM up to 0.23.1rc0. It has been declared as problematic. This vulnerability affects unknown code of the component API Image Handler. Executing a manipulation of the argument error.message can lead to log file. This vulnerability is handled as CVE-2026-54236. The attack can be executed remotely. There is not any exploit available. It is recommended to upgrade the affected component.

Detailsinfo

A vulnerability classified as problematic was found in vLLM up to 0.23.1rc0. This vulnerability affects some unknown processing of the component API Image Handler. The manipulation of the argument error.message with an unknown input leads to a log file vulnerability. The CWE definition for the vulnerability is CWE-532. Information written to log files can be of a sensitive nature and give valuable guidance to an attacker or expose sensitive user information. As an impact it is known to affect confidentiality. CVE summarizes:

vLLM is an inference and serving engine for large language models (LLMs). Prior to 0.23.1rc0, the fix for CVE-2026-22778, which introduced a sanitize_message helper that strips object-repr memory addresses from error messages before they reach the client, is incomplete: several response paths echo str(exc) directly to clients without calling sanitize_message. The unsanitized sites include the Anthropic API router in vllm/entrypoints/anthropic/api_router.py (the POST /v1/messages and POST /v1/messages/count_tokens handlers), the Server-Sent Events streaming converter in vllm/entrypoints/anthropic/serving.py, and the realtime speech-to-text WebSocket in vllm/entrypoints/speech_to_text/realtime/connection.py. These paths catch the exception inside the route coroutine and construct the JSONResponse themselves, bypassing the sanitizing global FastAPI exception handler, and WebSocket frames do not traverse that handler chain at all. Using the same primitive as the parent issue, an unauthenticated attacker can send malformed image bytes through the Anthropic Messages API image content parts so that PIL.Image.open raises an UnidentifiedImageError whose message contains the BytesIO object repr, leaking the heap memory address verbatim in the error.message field of the response body. This vulnerability is fixed in 0.23.1rc0.

The advisory is available at github.com. This vulnerability was named CVE-2026-54236 since 06/12/2026. The exploitation appears to be easy. The attack can be initiated remotely. No form of authentication is required for a successful exploitation. Technical details are known, but there is no available exploit. This vulnerability is assigned to T1592 by the MITRE ATT&CK project.

Upgrading to version 0.23.1rc0 eliminates this vulnerability. Applying the patch 94923629729381d7f7c9efde72071a2441f7fd82 is able to eliminate this problem. The bugfix is ready for download at github.com. The best possible mitigation is suggested to be upgrading to the latest version.

If you want to get best quality of vulnerability data, you may have to visit VulDB.

Productinfo

Name

• vLLM

Version

• 0.23.1rc0

License

• open-source

CPE 2.3info

• 🔒

CPE 2.2info

• 🔒

CVSSv4info

CVSSv3info

CVSSv2info

Exploitinginfo

Threat Intelligenceinfo

Countermeasuresinfo

Timelineinfo

Sourcesinfo

Entryinfo

If you want to get best quality of vulnerability data, you may have to visit VulDB.

Discussion