CVE-2026-126395: vLLM 0.14.0rc1 before 0.31.0 Heap Address Disclosure via sanitize_message()
vLLM from 0.14.0rc1 before 0.31.0 contains an information disclosure vulnerability in sanitizemessage() whose lowercase-only regex fails to scrub uppercase PIL image addresses. Unauthenticated attackers can send a crafted PNG to endpoints such as /tokenize, triggering HF-processor errors that leak live server heap addresses useful for ASLR bypass.
Affected Software
Event History
Frequently Asked Questions
Which deployments are exposed to this issue?
vLLM versions from 0.14.0rc1 through versions before 0.31.0 are affected. Exposure requires an endpoint that processes attacker-supplied PNG content, such as /tokenize, and can trigger an HF-processor error.
Does exploitation require authentication or user interaction?
No. An unauthenticated attacker can send a crafted PNG request to a vulnerable endpoint; no user interaction is required.
What information can an attacker obtain?
The error handling can disclose live server heap addresses because sanitize_message() does not remove uppercase PIL image addresses. These addresses may be useful for bypassing ASLR.