CVE-2026-115913Medium· 6.5▾ SunlitvLLM is an inference and serving engine for large language models. Prior to 0.32.0, the Phi-4 multimodal audio encoder in phi4mm_audio.py and phi4mm_utils.py uses eager attention by default, allowing a remote client to submit one legal l…
▾ Sunlit zone — Low / medium · no exploitation signal
impact 35.8 · likelihood 0 · exploitation 0
Need a working PoC? Pro members can cast a request and our team develops one — it lands right here.
vLLM is an inference and serving engine for large language models. Prior to 0.32.0, the Phi-4 multimodal audio encoder in phi4mm_audio.py and phi4mm_utils.py uses eager attention by default, allowing a remote client to submit one legal long audio clip that causes MultiHeadedAttention.forward_attention to materialize and retain full attention matrices in self.attn. The scaled dot-product attention branch does not materialize these matrices. The resulting quadratic GPU-memory use can cause an out-of-memory condition and terminate EngineCore for all users. This issue is fixed in version 0.32.0.
vllm < 0.32.0Refer to the linked advisories for vendor-supplied fixes and affected version ranges.
Connected by shared product, vendor, weakness, or advisory.
CVE-2026-115911Medium· 6.5vLLM: DoS via unbounded chat_template Jinja2 rendering exhausts the renderer thread pool
CVE-2026-115918Medium· 6.5vLLM: Responses API bypasses trust_request_chat_template for chat_template_kwargs
CVE-2026-115912High· 7.5vLLM: Public multimodal SHM handle forgery reaches worker-side `pickle.loads()` with insecure serialization disabled
CVE-2026-115916Medium· 6.5vLLM: Out-of-vocabulary prompt token id passes input validation and triggers a CUDA device-side assert (engine crash / DoS)
CVE-2026-115917Low· 3.4vLLM: Permissive default CORS origins allow cross-origin requests to the vLLM HTTP server
CVE-2026-115915High· 8.1vLLM: Untrusted EPS inputs can trigger denial of service or Ghostscript RCE