| Medium | /v1/systemone bypasses the --trust-request-chat-template gateGHSA-xgh8-qwjh-q853 | 2026-11-18 | 0.34.0 |
| Medium | Thinking-budget sparse state aliases onto a co-batched request during a mixed `SWAP`, forcing the reasoning-end token into another request'sGHSA-fpch-3xrf-7q96 | 2026-11-09 | 0.28.0 |
| Medium | Cross-user token data leakage in async scheduling + Pipeline ParallelismGHSA-v97p-wpxp-px84 | 2026-11-09 | 0.21.0 |
| Medium | Response amplification in OpenAI validation error handlerGHSA-mwgq-9582-w2vf | 2026-11-07 | 0.30.0 |
| High | Missing request-wide media budget allows single-request frontend resource exhaustionGHSA-5r3c-5x4v-8gjh | 2026-11-02 | 0.33.0 |
| High | Unauthenticated EngineCore crash via out-of-range prompt token id under skip_tokenizer_init (incomplete fix of GHSA-25q3-v2hm-8vpf)GHSA-q9v9-qpjp-fc24 | 2026-11-02 | 0.33.0 |
| Medium | vLLM Speculators Config verifier.name_or_path Element Directory Traversal Information Disclosure VulnerabilityGHSA-qrj2-j723-m6w8 | 2026-11-02 | 0.33.0 |
| Medium | vLLM: DoS via unbounded chat_template Jinja2 rendering exhausts the renderer thread poolCVE-2026-115911 · GHSA-542m-9234-8gvf | 2026-10-28 | 0.32.0 |
| High | vLLM: Public multimodal SHM handle forgery reaches worker-side `pickle.loads()` with insecure serialization disabledCVE-2026-115912 · GHSA-6757-hggm-hp2w | 2026-10-28 | 0.32.0 |
| Medium | vLLM: Phi-4 multimodal audio encoder materializes and retains full attention matrices: a single max-length audio OOM-crashes the vLLM engineCVE-2026-115913 · GHSA-8397-8qf6-f2gm | 2026-10-28 | 0.32.0 |
| Medium | vLLM DeepSeek-OCR image processing vulnerable to denial of serviceCVE-2026-115914 · GHSA-gq5q-89c7-vhvx | 2026-10-28 | 0.32.0 |
| High | vLLM: Untrusted EPS inputs can trigger denial of service or Ghostscript RCECVE-2026-115915 · GHSA-h3rc-j82p-2c3h | 2026-10-28 | 0.32.0 |
| Medium | vLLM: Out-of-vocabulary prompt token id passes input validation and triggers a CUDA device-side assert (engine crash / DoS)CVE-2026-115916 · GHSA-q97c-p368-75c4 | 2026-10-28 | 0.32.0 |
| Low | vLLM: Permissive default CORS origins allow cross-origin requests to the vLLM HTTP serverCVE-2026-115917 · GHSA-vvpg-rrww-wpmh | 2026-10-28 | 0.32.0 |
| Medium | vLLM: Responses API bypasses trust_request_chat_template for chat_template_kwargsCVE-2026-115918 · GHSA-x567-49g5-p4cf | 2026-10-28 | 0.32.0 |
| Medium | vLLM denial of service via prompt embeds on XD-RoPE models (incomplete fix of CVE-2026-55514)GHSA-5q78-j82c-2vr2 | 2026-10-27 | 0.30.0 |
| Medium | vLLM OpenCVDynamicOpenPanguVideoBackend: request-level num_frames without a ceiling — unauthenticated memory/CPU exhaustion (OOM-kill) on anGHSA-894j-wg8m-63h8 | 2026-10-27 | 0.31.0 |
| Medium | Quadratic Cohere2 Vision preprocessing DoS via request-controlled max_patchesGHSA-9g3c-x747-c6wr | 2026-10-27 | 0.31.0 |
| Medium | vllm p2p never connected zmq_exhaustionCVE-2026-94624 · GHSA-frvg-qw88-hww9 | 2026-10-27 | 0.32.0 |
| High | vLLM: Request-controlled mm_processor_kwargs.code_revision allows remote code executionCVE-2026-114053 · GHSA-h3rc-6mm3-gc2m | 2026-10-27 | 0.31.0 |
| Medium | sanitize_message() address regex is lowercase-only and does not match PIL's uppercase __repr__, leaking live heap addresses to unauthenticatGHSA-m9fg-w9qh-67pc | 2026-10-27 | 0.31.0 |
| Medium | Incomplete CVE-2026-34755 remediation can be bypassed by request-level num_framesGHSA-vxqj-p4gw-9h4c | 2026-10-27 | 0.31.0 |
| Medium | MiniMax M3 video preprocessing permits request-driven CPU resource exhaustion through min_pixelsGHSA-w4cg-mhc7-mgw3 | 2026-10-27 | 0.31.0 |
| High | vLLM from 0.20.1 before 0.29.0 contains an improper input validation vulnerability that allows remote attackers to crash the engine by supplying out-of-vocabulary stoptokenids.CVE-2026-116181 | 2026-10-25 | 0.29.0 |
| High | vLLM from 0.8.5 before 0.31.0 has a denial of service flaw.CVE-2026-116180 | 2026-10-25 | 0.31.0 |
| High | vLLM from 0.18.0 before 0.31.0 contains a resource exhaustion vulnerability that allows API clients to bypass the video/jpeg frame-count cap by overriding numframes through mediaiokwargs.CVE-2026-116179 | 2026-10-25 | 0.31.0 |
| High | vLLM from 0.20.0 before 0.30.0 contains a denial of service vulnerability that allows remote attackers to crash the engine by sending inconsistent multimodal metadata to the disaggregated /inference/v1/generate endpoint.CVE-2026-116178 | 2026-10-25 | 0.30.0 |
| High | vLLM from 0.10.2 before 0.25.0 contains an uncontrolled resource consumption vulnerability that allows unauthenticated attackers to exhaust memory because the Nemotron-VL processors globally set Image.MAXIMAGEPIXELS to None.CVE-2026-116177 | 2026-10-25 | 0.25.0 |
| High | vLLM from 0.9.2 before 0.24.0 contains a server-side request forgery vulnerability that allows API callers to force outbound connections by embedding a decode address in the X-Request-Id header.CVE-2026-116176 | 2026-10-25 | 0.24.0 |
| High | vLLM from 0.19.0 before 0.24.0 contains a decompression bomb vulnerability because the AudioMediaIO multimodal path calls loadaudio without the maxdurations decoded-duration limit.CVE-2026-116175 | 2026-10-25 | 0.24.0 |
| High | vLLM from 0.22.0 before 0.24.0 contains an improper input validation vulnerability in the Rust gRPC Generate and GenerateStream service that forwards unchecked prompt token IDs to EngineCore.CVE-2026-116174 | 2026-10-25 | 0.24.0 |
| High | vLLM from 0.22.0 before 0.24.0 contains a denial of service vulnerability that allows remote attackers to terminate EngineCore by bypassing logprobtokenids limits in the Rust gRPC frontend.CVE-2026-116173 | 2026-10-25 | 0.24.0 |
| High | vLLM from 0.22.0 before 0.24.0 contains an improper input validation vulnerability in the Rust gRPC converter that allows remote attackers to bypass maxlogprobs limits via CandidateTokens.topn.CVE-2026-116172 | 2026-10-25 | 0.24.0 |
| High | vLLM : DoS via unvalidated mm_processor_kwargs in multimodal image processors (Isaac, InternVL, H2OVL, etc.)CVE-2026-114055 · GHSA-gx7p-2j49-hfq4 | 2026-10-23 | 0.31.0 |
| Medium | vLLM: Forged multimodal UUID + P0/P1 processor-cache eviction drift: unauthenticated remote crash of the whole vLLM engineCVE-2026-114050 · GHSA-p92p-rxj5-7p2x | 2026-10-23 | 0.31.0 |
| High | vLLM before 0.31.0 contains a cache aliasing vulnerability that allows authenticated attackers to collide shared cache keys by supplying caller-controlled values such as EXIF ImageID or cachesalt.CVE-2026-114054 | 2026-10-21 | 0.31.0 |
| High | vLLM before 0.31.0 skips the mintokens <= maxtokens check for /inference/v1/generate requests with a null maxtokens, so the EngineCore input thread later dies on an uncaught VLLMValidationError while /health still returns 200.CVE-2026-114051 | 2026-10-21 | 0.31.0 |
| Medium | vLLM before 0.31.0 contains a cross-request mask leak in InputBatch.condense() that allows remote attackers to constrain other clients' outputs by leaving stale allowedtokenids rows.CVE-2026-114049 | 2026-10-21 | 0.31.0 |
| Medium | Remote DoS in disaggregated multimodal /inference/v1/generate via malicious kwargs_data / mm_placeholdersGHSA-34qw-q2h2-jp3m | 2026-10-09 | 0.30.0 |
| High | Frame-count cap bypass in VideoMediaIO.load_base64 for video/jpeg via media_io_kwargs.video.num_framesGHSA-4m6g-h9mv-wqjc | 2026-10-09 | 0.31.0 |
| Medium | Rust gRPC output_candidates.token_ids bypasses logprob_token_ids limits and can terminate EngineCoreGHSA-6pgw-f3pv-4h8q | 2026-10-09 | 0.24.0 |
| High | Remote DoS in vLLM via allowed_token_ids Speculative-Decoding Metadata MismatchGHSA-8c65-hq7q-r7jm | 2026-10-09 | 0.18.0 |
| High | Nemotron-VL processors disable Pillow decompression-bomb guard globally, enabling unauthenticated DoSGHSA-mm38-5g96-7j6x | 2026-10-09 | 0.25.0 |
| Medium | Multimodal chat audio decoding bypasses the decompression-bomb duration limitGHSA-r45p-7qwx-qwjw | 2026-10-09 | 0.24.0 |
| High | Cross-tenant DoS via uninitialized LogprobsTensors with prompt_logprobsGHSA-rhcx-5729-88vg | 2026-10-09 | 0.31.0 |
| Medium | Rust gRPC CandidateTokens.top_n bypasses max_logprobs and can terminate EngineCoreGHSA-rwfw-xvg3-3937 | 2026-10-09 | 0.24.0 |
| High | SSRF via X-Request-Id header drives P2pNcclConnector outbound ZMQ+NCCL pair, leaking KV-cache tensorsGHSA-vfp2-c8pq-v6h6 | 2026-10-09 | 0.24.0 |
| Medium | Rust gRPC Generate accepts out-of-vocabulary prompt token IDs and can terminate EngineCoreGHSA-w2v2-5662-8v5q | 2026-10-09 | 0.24.0 |
| Medium | Unvalidated stop_token_ids causes engine crash (DoS) and out-of-bounds GPU writeGHSA-w9r6-29q4-pvpr | 2026-10-09 | 0.29.0 |
| Low | A security flaw has been discovered in vllm-project vLLM up to 0.31.0.CVE-2026-105922 | 2026-10-06 | See advisory |
| Medium | An oversized min_tokens in one request hangs the vLLM engine while /health stays greenGHSA-4xqp-c3mv-qff7 | 2026-10-06 | 0.31.0 |
| Low | V1 InputBatch.condense leaves a stale allowed_token_ids mask on recycled batch rowsGHSA-6cxc-2vcg-w5qc | 2026-10-06 | 0.31.0 |
| High | MOSS-Audio request-controlled processor cache permits remote memory exhaustionCVE-2026-114052 · GHSA-823j-m4cj-hmjf | 2026-10-06 | 0.31.0 |
| Low | vLLM: Harmony tool continuations drop `cache_salt` — restoring a cross-tenant prefix-cache membership oracleCVE-2026-105752 · GHSA-935w-9g4m-p28p | 2026-10-06 | 0.30.0 |
| Medium | vLLM: Mirrored multimodal IPC caches desync after a rejected request — a later request reusing the same media hash trips a receiver assertioCVE-2026-105753 · GHSA-ph3r-5jfg-f84f | 2026-10-06 | 0.28.0 |
| Medium | Caller-controllable values become or collide with shared cache keys, enabling cross-request cache aliasing (2 sites: multimodal EXIF hash anGHSA-rh6f-3x46-j33q | 2026-10-06 | 0.31.0 |
| Medium | vLLM is an inference and serving engine for large language models.CVE-2026-105759 | 2026-10-05 | 0.30.0 |
| Medium | vLLM: Loose `cache_salt` validation lets a single request kill EngineCore on LMCache-MP deployments — uncaught downstream `ValueError` deniaCVE-2026-105756 · GHSA-2823-qmq8-rwvj | 2026-10-05 | 0.30.0 |
| Medium | vLLM: Flash late-interaction scoring caches query embeddings under a caller-controlled request id — cross-request integrity break and induceCVE-2026-105755 · GHSA-2phq-3phc-84px | 2026-10-05 | 0.30.0 |
| Medium | vLLM: GLMGA video sampling permits request-driven CPU and memory exhaustionCVE-2026-105760 · GHSA-58v5-2m8f-94pr | 2026-10-05 | 0.30.0 |
| Medium | vLLM: Structured-output request errors escape the request boundary and terminate the shared EngineCore — engine-fatal denial of service (3 sCVE-2026-105757 · GHSA-85xf-c7hm-whqw | 2026-10-05 | 0.30.0 |
| Medium | vLLM: Scale-out disaggregated multimodal transport trusts caller-supplied featuresCVE-2026-105754 · GHSA-ph72-cqr5-qpp7 | 2026-10-05 | 0.30.0 |
| Medium | vLLM: Qwen2-VL / Qwen3-VL video samplers bound on request-controlled max_frames, which the num_frames ceiling does not reachCVE-2026-105758 · GHSA-x6mc-67gf-chw4 | 2026-10-05 | 0.30.0 |
| Medium | A flaw has been found in vllm-project vLLM up to 0.26.0.CVE-2026-103241 | 2026-09-30 | See advisory |
| Medium | Crafted IAMF audio upload reaches a PyAV/FFmpeg native heap overflow through the speech transcription path — denial of serviceGHSA-m52c-39rh-f3gp | 2026-09-28 | 0.29.0 |
| High | vLLM before 0.29.0 accepts user-controlled stoptokenids on the OpenAI-compatible POST /v1/completions and POST /v1/chat/completions endpoints but validates only that the values are integers, not that each token id is within the model…CVE-2026-100654 | 2026-09-26 | 0.29.0 |
| High | vLLM is an inference and serving engine for large language models.CVE-2026-100653 | 2026-09-26 | 0.28.0 |
| High | vLLM versions 0.22.0 through 0.23.0 fail to validate stoptokenids against vocabulary bounds in Rust HTTP and gRPC frontends, allowing out-of-vocabulary token IDs to reach MinTokensLogitsProcessor.CVE-2026-100652 | 2026-09-26 | 0.24.0 |
| High | vLLM before 0.29.0 fails to enforce decoder prompt-length validation on the disaggregated serving endpoint /inference/v1/generate.CVE-2026-100651 | 2026-09-26 | 0.29.0 |
| High | vLLM through 0.29.0 fetches and fully materializes remote or inline media before enforcing its documented media controls (the VLLMMAXAUDIOCLIPFILESIZEMB compressed-audio size cap, default 25 MB, and the per-modality --limit-mm-per-prompt…CVE-2026-100650 | 2026-09-26 | See advisory |
| High | vLLM Mooncake connector through 0.29.0 fails to properly manage GPU KV cache block ownership when concurrent child requests share a single transfer ID in prefill/decode disaggregated deployments.CVE-2026-94627 | 2026-09-21 | See advisory |
| High | vLLM through 0.29.0 fails to validate the tpsize parameter in kvtransferparams on OpenAI-compatible completion endpoints, allowing attackers to allocate unbounded memory.CVE-2026-94626 | 2026-09-21 | See advisory |
| Medium | vLLM through 0.29.0 contains a resource exhaustion vulnerability in MooncakeConnector where rejected prefill requests create ownerless transfer placeholders that are never reclaimed.CVE-2026-94625 | 2026-09-21 | See advisory |
| High | vLLM through 0.29.0 contains a denial of service vulnerability in the NIXL connector's prefix caching implementation that fails to properly validate block counts across multi-prompt completion requests in prefill/decode disaggregated…CVE-2026-94623 | 2026-09-21 | See advisory |
| High | vLLM versions through 0.29.0 contain a denial of service vulnerability in the NIXL connector's metadata handling for prefill/decode disaggregated deployments.CVE-2026-94622 | 2026-09-21 | See advisory |
| Low | vLLM through 0.29.0 fails to properly validate badwords token indices against the model's generation output width in SamplingParams.updatefrom_tokenizer().CVE-2026-93989 | 2026-09-19 | See advisory |
| Medium | vLLM through 0.29.0 contains a memory corruption vulnerability in the Triton bincountkernel where prompt token IDs index the penalty prompt-presence bitset without bounds checking against vocabulary size.CVE-2026-93841 | 2026-09-18 | See advisory |
| Medium | vLLM before 0.29.0 validates allowedtokenids against tokenizer length instead of model output logits width in SamplingParams.validateallowedtokenids().CVE-2026-93840 | 2026-09-18 | 0.29.0 |
| High | vLLM through 0.29.0 fails to properly clean up decode-side metadata for rejected inference requests in prefill/decode disaggregated deployments.CVE-2026-93436 | 2026-09-17 | See advisory |
| Medium | vLLM: Request-selected PyNvVideoCodec GPU decode bypasses static VRAM reservationCVE-2026-69147 · GHSA-8pw2-6jv3-mj5j | 2026-09-17 | 0.28.0 |
| Medium | vLLM: Unauthenticated audio decompression-bomb DoS in /v1/chat/completionsCVE-2026-57173 · GHSA-hcwq-8wjf-3gcr | 2026-09-16 | 0.24.0 |
| Medium | Disaggregated generate skips decoder prompt-length validation for some multimodal processorsGHSA-3mqx-f33v-vgp9 | 2026-09-12 | 0.29.0 |
| Medium | Incomplete artifact pin propagation in FunAudioChat and Tarsier2GHSA-hhv2-872h-628q | 2026-09-12 | 0.28.0 |
| Low | Sampler Subclass Counter Shadowing Bypasses PyNvVideoCodec Decoder Limits and GPU Memory AccountingCVE-2026-100649 · GHSA-j682-9xp5-rrf3 | 2026-09-12 | 0.29.0 |
| Medium | Uncontrolled resource consumption: multimodal chat audio decoding ignores `VLLM_MAX_AUDIO_CLIP_FILESIZE_MB`CVE-2026-100648 · GHSA-jcq2-4gch-5qhf | 2026-09-12 | 0.29.0 |
| Medium | Remote media is fetched and fully materialized before the documented media size / item limits are enforced — remote pre-inference memory andGHSA-p6g9-7v3x-m8mv | 2026-09-12 | See advisory |
| Medium | Rust HTTP/gRPC stop_token_ids bypass Python vocab-bound fix and can terminate EngineCoreGHSA-qff2-492f-9fm4 | 2026-09-12 | 0.24.0 |
| Medium | Out-of-range `stop_token_ids` with `min_tokens` can kill vLLM EngineCoreGHSA-v5gm-qgmv-gc6c | 2026-09-12 | 0.29.0 |
| Medium | vLLM DoS via unbounded `cache_salt` length: multi-hundred-MB salt stalls the single EngineCore scheduler thread (CPU exhaustion)CVE-2026-100647 · GHSA-wpww-v874-ph2p | 2026-09-12 | 0.29.0 |
| Medium | vLLM: SSRF + arbitrary local file read in MiMoV2OmniMultiModalProcessor `_fetch_image` and audio loader bypass MediaConnector protectionsCVE-2026-73560 · GHSA-4hhp-h66f-j5j7 | 2026-09-08 | 0.26.0 |
| Medium | vLLM: Cross-User Data Leak VulnerabilityCVE-2026-73558 · GHSA-7m6h-x95x-82q5 | 2026-09-08 | 0.27.0 |
| Medium | vLLM: ReDoS via structured_outputs.regex in the lm-format-enforcer backend (no compile timeout) — missed sibling of GHSA-rwxx-mrjm-wc2mCVE-2026-73556 · GHSA-48jh-3gj7-fg8v | 2026-09-04 | 0.26.0 |
| Medium | vLLM: Derender endpoints decode caller-supplied GenerateResponse token IDs without output boundsCVE-2026-71486 · GHSA-8737-qx52-hjff | 2026-09-04 | 0.26.0 |
| Medium | vLLM: Unauthenticated Internal Path and Username Disclosure via Validation Error MessagesCVE-2026-73555 · GHSA-hwrm-c4cx-rf4j | 2026-09-04 | 0.26.0 |
| Medium | vLLM: Incomplete CVE-2025-62164 remediation can be bypassed by concurrent prompt partsCVE-2026-73557 · GHSA-pr7f-p5mw-fc87 | 2026-09-04 | 0.26.0 |
| High | Denial of service: a negative token id in `/v1/embeddings` or `/pooling` input kills the vLLM engine via a CUDA device-side assertionCVE-2026-93592 · GHSA-25q3-v2hm-8vpf | 2026-09-03 | 0.28.0 |
| High | vLLM up to and including 0.17.0 allows remote attackers to cause a Denial of Service via memory exhaustion.CVE-2026-37237 | 2026-08-28 | See advisory |
| High | LlavaOnevision2 processor loader executes attacker model code with `trust_remote_code=False` (inert `trust_remote_code` kwarg to `transformeCVE-2026-90553 · GHSA-3c86-2m5g-59q7 | 2026-08-28 | 0.28.0 |
| Medium | Denial of Service NanoNemoTronVL Video Audio Extraction BombCVE-2026-90554 · GHSA-936p-m5pv-vvjf | 2026-08-28 | 0.28.0 |
| Medium | Speech-to-text audio decode duration limit bypass via forged header sample rateCVE-2026-90555 · GHSA-99f2-hwrc-gvq8 | 2026-08-28 | 0.28.0 |
| Medium | vLLM: Completion prompt lists fan out into unbounded engine requestsCVE-2026-73559 · GHSA-87x5-vmc3-756j | 2026-08-13 | 0.26.0 |
| Medium | vLLM Unauthenticated Requests Exploit DeepStream Backend Confusion for DoSCVE-2026-78684 · GHSA-cqm8-jxg6-fqfq | 2026-08-11 | 0.27.0 |
| High | vLLM denial of service via prompt embeds on M-RoPE modelsCVE-2026-55514 · GHSA-33cg-gxv8-3p8g | 2026-07-20 | 0.24.0 |
| Medium | vLLM: Processing differential in multi-channel audio downmixing enables hidden-input/moderation bypass for audio modelsCVE-2026-34760 · GHSA-6c4r-fmh3-7rh8 | 2026-07-17 | 0.18.0 |
| High | vLLM has Remote DoS via Invalid Recovered Token ReinjectionCVE-2026-54234 · GHSA-8wr5-jm2h-8r4f | 2026-07-17 | 0.24.0 |
| High | vLLM: ReDoS via structured_outputs.regex compiled without timeout in xgrammar and outlines backendsCVE-2026-55574 · GHSA-rwxx-mrjm-wc2m | 2026-07-17 | 0.24.0 |
| Medium | vLLM: Speech-to-text upload size limit is enforced after full UploadFile readCVE-2026-55646 · GHSA-v82g-2437-67m2 | 2026-07-17 | 0.24.0 |
| High | vLLM is an inference and serving engine for large language models (LLMs).CVE-2026-54232 | 2026-06-22 | 0.22.1 |
| Medium | vLLM versions >= 0.6.3 and < 0.9.0 contain multiple regular expression denial of service (ReDoS) vulnerabilities.CVE-2025-71379 | 2026-06-20 | 0.9.0 |
| Medium | vLLM: GGUF dequantize kernel int truncation exposes uninitialized GPU memory in multi-tenant servingCVE-2026-53923 · GHSA-5jv2-g5wq-cmr4 | 2026-06-17 | 0.24.0 |
| Medium | vLLM: OOM Denial of Service via Audio Decompression BombCVE-2026-54233 · GHSA-6pr9-rp53-2pmc | 2026-06-17 | 0.24.0 |
| Medium | vLLM: temperature=NaN and temperature=Infinity bypass validation and propagate to GPU kernelsCVE-2026-54235 · GHSA-7h4p-rffg-7823 | 2026-06-17 | 0.24.0 |
| Medium | vLLM: image EXIF Rotation & PNG tRNS Transparency Not Normalized, Causing Mismatch Between Model Input and ExpectationsCVE-2026-12491 · GHSA-8jr5-v98p-w75m | 2026-06-17 | 0.24.0 |
| Medium | vLLM: incomplete CVE-2026-22778 fix leaks PIL repr addresses via Anthropic routerCVE-2026-54236 · GHSA-hgg8-fqqc-vfmw | 2026-06-17 | 0.24.0 |
| Critical | vLLM: OpenAI auth bypassCVE-2026-48746 · GHSA-94f4-hr76-p5j6 | 2026-06-16 | 0.22.0 |
| High | vLLM: Security Check Bypass via assert Statement in Activation Function Loading Allows Arbitrary Code ExecutionCVE-2026-41523 · GHSA-q8gq-377p-jq3r | 2026-06-16 | 0.22.0 |
| High | vLLM is vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to unbounded frame count processing in the `VideoMediaIO.loaCVE-2026-5497 · GHSA-wcwg-c5fc-9vrc | 2026-06-11 | 0.19.0 |
| Medium | vLLM's Artifact Pin Decay allows pinned deployments to load unpinned code, weights, and processorsCVE-2026-47155 · GHSA-3ww4-5jv9-j5gm | 2026-06-10 | 0.22.0 |
| Medium | vllm has Improper Resource Shutdown or ReleaseCVE-2026-9540 · GHSA-98f3-hwg4-4rf7 | 2026-05-26 | See advisory |
| Medium | vLLM: extract_hidden_states speculative decoding crashes server on any request with penalty parametersCVE-2026-44223 · GHSA-83vm-p52w-f9pw | 2026-05-06 | 0.20.0 |
| Medium | vLLM Vulnerable to Remote DoS via Special-Token PlaceholdersCVE-2026-44222 · GHSA-hpv8-x276-m59f | 2026-05-05 | 0.20.0 |
| Low | vLLM makes Use of Uninitialized ResourceCVE-2026-7141 · GHSA-x368-4g9h-fvv4 | 2026-04-27 | 0.19.1 |
| Medium | vLLM: Unauthenticated OOM Denial of Service via Unbounded `n` Parameter in OpenAI API ServerCVE-2026-34756 · GHSA-3mwp-wvh9-7528 | 2026-04-03 | 0.19.0 |
| Medium | vLLM: Server-Side Request Forgery (SSRF) in `download_bytes_from_url `CVE-2026-34753 · GHSA-pf3h-qjgv-vcpr | 2026-04-03 | 0.19.0 |
| Medium | vLLM: Denial of Service via Unbounded Frame Count in video/jpeg Base64 ProcessingCVE-2026-34755 · GHSA-pq5c-rjhq-qp7p | 2026-04-03 | 0.19.0 |
| High | vLLM has Hardcoded Trust Override in Model Files Enables RCE Despite Explicit User Opt-OutCVE-2026-27893 · GHSA-7972-pg2x-xr59 | 2026-03-27 | 0.18.0 |
| Medium | vLLM has SSRF Protection BypassCVE-2026-25960 · GHSA-v359-jj2v-j536 | 2026-03-09 | 0.17.0 |
| Critical | vLLM has RCE In Video ProcessingCVE-2026-22778 · GHSA-4r2x-xpjr-7cvv | 2026-02-02 | 0.14.1 |
| High | vLLM vulnerable to Server-Side Request Forgery (SSRF) through MediaConnectorCVE-2026-24779 · GHSA-qh4c-xf7m-gxfc | 2026-01-28 | 0.14.1 |
| High | vLLM affected by RCE via auto_map dynamic module loading during model initializationCVE-2026-22807 · GHSA-2pc9-4j83-qjmr | 2026-01-21 | 0.14.0 |
| Medium | vLLM is vulnerable to DoS in Idefics3 vision models via image payload with ambiguous dimensionsCVE-2026-22773 · GHSA-grg2-63fw-f2qr | 2026-01-13 | 0.12.0 |
| High | vLLM introduced enhanced protection for CVE-2025-62164CVE-2026-56340 · GHSA-mcmc-2m55-j8jj | 2026-01-08 | 0.13.0 |
| High | vLLM vulnerable to remote code execution via transformers_utils/get_configCVE-2025-66448 · GHSA-8fr4-5q9j-m8gm | 2025-12-02 | 0.11.1 |
| Medium | vLLM vulnerable to DoS via large Chat Completion or Tokenization requests with specially crafted `chat_template_kwargs`CVE-2025-62426 · GHSA-69j4-grxj-j64p | 2025-11-20 | 0.11.1 |
| High | vLLM deserialization vulnerability leading to DoS and potential RCECVE-2025-62164 · GHSA-mrw7-hf4f-83pf | 2025-11-20 | 0.11.1 |
| High | vLLM vulnerable to DoS with incorrect shape of multimodal embedding inputsCVE-2025-62372 · GHSA-pmqf-x6x8-p7qw | 2025-11-20 | 0.11.1 |
| High | vLLM is vulnerable to Server-Side Request Forgery (SSRF) through `MediaConnector` classCVE-2025-6242 · GHSA-3f6c-7fw2-ppm4 | 2025-10-07 | 0.11.0 |
| Medium | vLLM: Resource-Exhaustion (DoS) through Malicious Jinja Template in OpenAI-Compatible ServerCVE-2025-61620 · GHSA-6fvq-23cw-5628 | 2025-10-07 | 0.11.0 |
| High | vLLM is vulnerable to timing attack at bearer authCVE-2025-59425 · GHSA-wr9h-g72x-mwhm | 2025-10-07 | 0.11.0 |
| High | vLLM has remote code execution vulnerability in the tool call parser for Qwen3-CoderCVE-2025-9141 · GHSA-79j6-g2m3-jgfw | 2025-08-21 | 0.10.1.1 |
| High | vllm API endpoints vulnerable to Denial of Service AttacksCVE-2025-48956 · GHSA-rxc4-3w6r-4v47 | 2025-08-21 | 0.10.1.1 |
| Low | Potential Timing Side-Channel Vulnerability in vLLM’s Chunk-Based Prefix CachingCVE-2025-46570 · GHSA-4qjh-9fv9-r85r | 2025-05-28 | 0.9.0 |
| Medium | vLLM DOS: Remotely kill vllm over http with invalid JSON schemaCVE-2025-48942 · GHSA-6qc9-v4r8-22xg | 2025-05-28 | 0.9.0 |
| Medium | vLLM allows clients to crash the openai server with invalid regexCVE-2025-48943 · GHSA-9hcf-v7m4-6m2j | 2025-05-28 | 0.9.0 |
| Medium | vLLM has a Weakness in MultiModalHasher Image Hashing ImplementationCVE-2025-46722 · GHSA-c65p-x677-fgj6 | 2025-05-28 | 0.9.0 |
| Medium | vLLM vulnerable to Regular Expression Denial of ServiceGHSA-j828-28rj-hfhp | 2025-05-28 | 0.9.0 |
| Medium | vLLM Tool Schema allows DoS via Malformed pattern and type FieldsCVE-2025-48944 · GHSA-vrq3-r879-7m65 | 2025-05-28 | 0.9.0 |
| Medium | vLLM has a Regular Expression Denial of Service (ReDoS, Exponential Complexity) Vulnerability in `pythonic_tool_parser.py`CVE-2025-48887 · GHSA-w6q7-j642-7c25 | 2025-05-28 | 0.9.0 |
| Critical | vLLM Allows Remote Code Execution via PyNcclPipe Communication ServiceCVE-2025-47277 · GHSA-hjq4-87xh-g4fv | 2025-05-20 | 0.8.5 |
| High | Remote Code Execution Vulnerability in vLLM Multi-Node Cluster ConfigurationCVE-2025-30165 · GHSA-9pcc-gvx5-r5wm | 2025-05-06 | 0.10.0 |
| High | Data exposure via ZeroMQ on multi-node vLLM deploymentCVE-2025-30202 · GHSA-9f8f-2vmf-885j | 2025-04-29 | 0.8.5 |
| Critical | vLLM Vulnerable to Remote Code Execution via Mooncake IntegrationCVE-2025-32444 · GHSA-hj4w-hm2g-p6w5 | 2025-04-29 | 0.8.5 |
| Medium | vLLM: Quadratic Time Complexity in Input Token Processing leads to denial of serviceCVE-2025-46560 · GHSA-vc6m-hm49-g9qg | 2025-04-29 | 0.8.5 |
| Critical | CVE-2025-24357 Malicious model remote code execution fix bypass with PyTorch < 2.6.0GHSA-ggpf-24jw-3fcw | 2025-04-23 | 0.8.0 |
| Medium | vLLM vulnerable to Denial of Service by abusing xgrammar cacheGHSA-hf3c-wxg2-49q9 | 2025-04-15 | 0.8.4 |
| Critical | vLLM Deserialization of Untrusted Data vulnerabilityCVE-2024-11041 · GHSA-5vqr-wprc-cpp7 | 2025-03-20 | See advisory |
| Critical | vLLM allows Remote Code Execution by Pickle Deserialization via AsyncEngineRPCServer() RPC server entrypointsCVE-2024-9053 · GHSA-cj47-qj6g-x7r4 | 2025-03-20 | See advisory |
| Critical | vLLM deserialization vulnerability in vllm.distributed.GroupCoordinator.recv_objectCVE-2024-9052 · GHSA-pgr7-mhp5-fgjp | 2025-03-20 | See advisory |
| Medium | vLLM denial of service via outlines unbounded cache on diskCVE-2025-29770 · GHSA-mgrm-fgjv-mhv8 | 2025-03-19 | 0.8.0 |
| Critical | vLLM Allows Remote Code Execution via Mooncake IntegrationCVE-2025-29783 · GHSA-x3m8-f7g5-qhm7 | 2025-03-19 | 0.8.0 |
| Low | vLLM uses Python 3.12 built-in hash() which leads to predictable hash collisions in prefix cacheCVE-2025-25183 · GHSA-rm76-4mrf-v9r8 | 2025-02-06 | 0.7.2 |
| High | vllm: Malicious model to RCE by torch.load in hf_model_weights_iteratorCVE-2025-24357 · GHSA-rh4j-5rhw-hr54 | 2025-01-27 | 0.7.0 |
| High | vLLM denial of service vulnerabilityCVE-2024-8768 · GHSA-w2r7-9579-27hf | 2024-09-17 | 0.5.5 |
| Medium | vLLM Denial of Service via the best_of parameterCVE-2024-8939 · GHSA-wc36-9694-f9rf | 2024-09-17 | See advisory |