Skip to contentVllm vulnerabilities
Newest
- vllmvLLM through 0.29.0 Cross-Request Logits Corruption via bad_wordsCVE-2026-93989Fix available2.3 Low
- vllmvLLM through 0.29.0 Adjacent-Request Sampler State Corruption via Unvalidated Prompt Token IDsCVE-2026-93841Fix available6.3 Medium
- vllmvLLM before 0.29.0 Cross-Request Logits Corruption via allowed_token_idsCVE-2026-93840Fix available6.3 Medium
- vllmvLLM before 0.28.0 Denial of Service via negative token IDCVE-2026-93592Fix available8.7 High
- vllmvLLM through 0.29.0 Memory Exhaustion via Rejected RequestsCVE-2026-93436Fix available8.7 High
- vllmvLLM: Request-selected PyNvVideoCodec GPU decode bypasses static VRAM reservationCVE-2026-69147Fix available6.5 Medium
- vllmvLLM: Unauthenticated audio decompression-bomb DoS in /v1/chat/completionsCVE-2026-57173Fix available6.5 Medium
- vllmvLLM before 0.28.0 Denial of Service via Audio HeaderCVE-2026-905557.1 High
- vllmvLLM before 0.28.0 Denial of Service via audio extractionCVE-2026-905546.9 Medium
- vllmvLLM before 0.28.0 Remote Code Execution via LlavaOnevision2 processorCVE-2026-90553Fix available8.5 High
- vllmvLLM up to and including 0.17.0 allows remote attackers to cause a Denial of Service via memory...CVE-2026-372377.5 High
- vllmvLLM before 0.27.0 Denial of Service via DeepStream BackendCVE-2026-78684Fix available6.9 Medium
- vllmvLLM: SSRF + arbitrary local file read in MiMoV2OmniMultiModalProcessor `_fetch_image` and audio loader bypass MediaConnector protectionsCVE-2026-73560Fix available6.5 Medium
- vllmvLLM: Derender endpoints decode caller-supplied GenerateResponse token IDs without output boundsCVE-2026-71486Fix available4.3 Medium
- vllmvLLM: Completion prompt lists fan out into unbounded engine requestsCVE-2026-73559Fix available6.5 Medium
- vllmvLLM: Cross-User Data Leak VulnerabilityCVE-2026-73558Fix available5.3 Medium
- vllmvLLM: Incomplete CVE-2025-62164 remediation can be bypassed by concurrent prompt partsCVE-2026-73557Fix available6.3 Medium
- vllmvLLM: ReDoS via structured_outputs.regex in the lm-format-enforcer backend (no compile timeout) — missed sibling of CVE-2026-55574CVE-2026-73556Fix available5.3 Medium
- vllmvLLM: Unauthenticated Internal Path and Username Disclosure via Validation Error MessagesCVE-2026-73555Fix available5.3 Medium
- vllmvLLM denial of service via prompt embeds on M-RoPE modelsCVE-2026-55514Fix available7.1 High
- vllmvLLM: ReDoS via structured_outputs.regex compiled without timeout in xgrammar and outlines backendsCVE-2026-55574Fix available8.7 High
- vllmvLLM: Remote DoS in vLLM via Invalid Recovered Token ReinjectionCVE-2026-54234Fix available7.5 High
- vllmvLLM speech-to-text endpoints allocate full upload before enforcing the audio file-size limitCVE-2026-55646Fix available6.5 Medium
- vllmvLLM: Dependency Confusion Vulnerability in vLLM DockerfileCVE-2026-54232Fix available8.8 High
- vllmvLLM: OOM Denial of Service via Audio Decompression BombCVE-2026-54233Fix available6.5 Medium
- vllmvLLM: incomplete CVE-2026-22778 fix leaks PIL repr addresses via Anthropic routerCVE-2026-54236Fix available5.3 Medium
- vllmvLLM GGUF Kernels: int64_t to int truncation of tensor dimensions causes GPU buffer overflowCVE-2026-53923Fix available5.3 Medium
- vllmvLLM: temperature=NaN and temperature=Infinity bypass validation and propagate to GPU kernelsCVE-2026-54235Fix available6.9 Medium
- vllm, red hat exploit intelligence, openshift lightspeed +12 more +14 morevLLM: OpenAI auth bypassCVE-2026-48746Fix available9.1 Critical
- vllm, red hat ai inference server, ai inference server 3.2 +5 more +7 morevLLM: Security Check Bypass via assert Statement in Activation Function Loading Allows Arbitrary Code ExecutionCVE-2026-41523Fix available7.5 High
- vllm, vllm-project vllm-project/vllm, red hat ai inference server +2 more +4 moreUnbounded Frame Count in video/jpeg Base64 Data URL Processing Leads to OOM DoS in vllm-project/vllmCVE-2026-5497Fix available7.5 High
- vllmvLLM: Artifact Pin Decay in vLLM allows pinned deployments to load unpinned code, weights, and processorsCVE-2026-47155Fix available6.5 Medium
- vllmvLLM: extract_hidden_states speculative decoding crashes server on any request with penalty parametersCVE-2026-44223Fix available6.5 Medium
- vllmvLLM: Remote DoS via Special-Token PlaceholdersCVE-2026-44222Fix available6.5 Medium
- vllmvLLM KV Block kv_cache_interface.py has_mamba_layers uninitialized resourceCVE-2026-7141Fix availableDisputed6.3 Medium
- vllm, red hat ai inference server, ai inference server 3.2 +3 more +5 morevLLM Affected by Denial of Service via Unbounded Frame Count in video/jpeg Base64 ProcessingCVE-2026-34755Fix available6.5 Medium
- vllmvLLM affected by Server-Side Request Forgery (SSRF) in `download_bytes_from_url `CVE-2026-34753Public exploitFix available5.4 Medium
- vllm, red hat ai inference server, ai inference server 3.2 +3 more +5 morevLLM Affected by Unauthenticated OOM Denial of Service via Unbounded `n` Parameter in OpenAI API ServerCVE-2026-34756Fix available6.5 Medium
- vllmvLLM: Downmix Implementation Differences as Attack Vectors Against Audio AI ModelsCVE-2026-34760Fix available5.9 Medium
- vllm, red hat ai inference server, ai inference server 3.2 +5 more +7 morevLLM's hardcoded trust_remote_code=True in NemotronVL and KimiK25 bypasses user security opt-outCVE-2026-27893Fix available8.8 High
- vllm, red hat ai inference server, enterprise linux ai (rhel ai) 3 +2 more +4 moreSSRF Protection Bypass in vLLMCVE-2026-25960Fix available7.1 High
- vllm, red hat ai inference server, ai inference server 3.2 +5 more +7 morevLLM leaks a heap address when PIL throws an errorCVE-2026-22778Public exploitFix availableEPSS 18.3%9.8 Critical
- vllm, red hat ai inference server, ai inference server 3.2 +5 more +7 morevLLM vulnerable to Server-Side Request Forgery (SSRF) in `MediaConnector`CVE-2026-24779Fix available7.1 High
- vllm, red hat ai inference server, ai inference server 3.2 +6 more +8 morevLLM affected by RCE via auto_map dynamic module loading during model initializationCVE-2026-22807Public exploitFix available8.8 High
- vllmvLLM is vulnerable to DoS in Idefics3 vision models via image payload with ambiguous dimensionsCVE-2026-22773Fix available6.5 Medium
- vllm, red hat ai inference server, enterprise linux ai (rhel ai) 3 +1 more +3 morevLLM - Denial of Service via Unvalidated Multimodal EmbeddingsCVE-2026-56340Fix available8.7 High
- vllmvLLM vulnerable to remote code execution via transformers_utils/get_configCVE-2025-66448Fix available7.1 High
- vllmvLLM vulnerable to DoS via large Chat Completion or Tokenization requests with specially crafted `chat_template_kwargs`CVE-2025-62426Fix available6.5 Medium
- vllmvLLM vulnerable to DoS with incorrect shape of multimodal embedding inputsCVE-2025-62372Fix available8.3 High
- vllmVLLM deserialization vulnerability leading to DoS and potential RCECVE-2025-62164Fix available8.8 High
- Show moreLoading…