{"kind": "product", "path": "vulns/pypi/vllm", "name": "vllm", "product": "vllm", "ecosystem": "PyPI", "items": [{"id": "CVE-2026-114055", "noindex": false, "published": "2026-10-21T00:36:31.227Z", "modified": "2026-10-21T15:50:46.285Z", "cvss": 8.7, "cvss_vector": "CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N", "severity": "high", "epss": 0.0042, "kev": false, "exploits": [], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-114055", "title": "vLLM 0.7.3 before 0.31.0 DoS via Unvalidated mm_processor_kwargs", "description": "vLLM from 0.7.3 before 0.31.0 contains an uncontrolled resource consumption vulnerability that allows remote clients to control multimodal processor parameters by supplying per-request mm_processor_kwargs. Attackers can set large image size, patch count, crop count, or frame count values in Isaac, InternVL, H2OVL, and other processors to exhaust memory and terminate the shared server.", "references": ["https://github.com/vllm-project/vllm/security/advisories/GHSA-gx7p-2j49-hfq4", "https://www.vulncheck.com/advisories/vllm-0.7.3-before-0.31.0-dos-via-unvalidated-mm-processor-kwargs", "https://nvd.nist.gov/vuln/detail/CVE-2026-114055"], "fixes": ["0.31.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}, {"id": "CVE-2026-114050", "noindex": false, "published": "2026-10-21T00:36:27.641Z", "modified": "2026-10-21T15:50:53.921Z", "cvss": 7.1, "cvss_vector": "CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N", "severity": "high", "epss": 0.00203, "kev": false, "exploits": [], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-114050", "title": "vLLM before 0.31.0 Remote Engine Crash via Forged Multimodal UUID", "description": "vLLM before 0.31.0 uses a client-supplied multimodal uuid verbatim as cache identity with no consistency check on cache hits, allowing engine crashes. Unauthenticated attackers can desynchronize the P0/P1 processor caches to trigger a fatal token-placeholder mismatch, killing the entire serving instance until restarted.", "references": ["https://github.com/vllm-project/vllm/security/advisories/GHSA-p92p-rxj5-7p2x", "https://www.vulncheck.com/advisories/vllm-before-0.31.0-remote-engine-crash-via-forged-multimodal-uuid", "https://nvd.nist.gov/vuln/detail/CVE-2026-114050"], "fixes": ["0.31.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}, {"id": "CVE-2026-105922", "noindex": false, "published": "2026-10-06T14:45:15.075Z", "modified": "2026-10-06T17:39:56.919Z", "cvss": 5.3, "cvss_vector": "CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:N/VA:L/SC:N/SI:N/SA:N/E:P", "severity": "medium", "epss": 0.00311, "kev": false, "exploits": [], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-105922", "title": "vllm-project vLLM Penalty utils.py get_token_bin_counts_and_mask denial of service", "description": "A security flaw has been discovered in vllm-project vLLM up to 0.31.0. This impacts the function get_token_bin_counts_and_mask of the file vllm/model_executor/layers/utils.py of the component Penalty Handler. Performing a manipulation results in denial of service. Remote exploitation of the attack is possible. The exploit has been released to the public and may be used for attacks. The project was informed of the problem early through an issue report but has not responded yet.", "references": ["https://vuldb.com/vuln/413896", "https://vuldb.com/vuln/413896/cti", "https://vuldb.com/cve/CVE-2026-105922"], "fixes": ["0.32.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}, {"id": "CVE-2026-105760", "noindex": false, "published": "2026-10-05T23:01:54.972Z", "modified": "2026-10-06T14:39:17.058Z", "cvss": 5.3, "cvss_vector": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:L", "severity": "medium", "epss": 0.00297, "kev": false, "exploits": ["public-poc"], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-105760", "title": "vLLM: GLMGA video sampling permits request-driven CPU and memory exhaustion", "description": "vLLM is an inference and serving engine for large language models. Prior to 0.30.0, a caller can use the request-level media_io_kwargs field to select the GLMGA video backend and supply large values for the fps and max_frames options without a strict work ceiling. GLMGA constructs and deduplicates an attacker-sized pre-decode frame-index list, allowing a compact request and tiny valid video to consume disproportionate CPU time and memory in the shared media-loading executor. This issue is fixed in version 0.30.0.", "references": ["https://github.com/vllm-project/vllm/security/advisories/GHSA-58v5-2m8f-94pr", "https://github.com/vllm-project/vllm/pull/54935", "https://github.com/vllm-project/vllm/commit/8b6de0eb9a09ef53f20cf06bd4d17ee264b9c2a7"], "fixes": ["0.30.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}, {"id": "CVE-2026-105759", "noindex": false, "published": "2026-10-05T22:58:01.362Z", "modified": "2026-10-06T17:56:44.425Z", "cvss": 5.9, "cvss_vector": "CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H", "severity": "medium", "epss": 0.00411, "kev": false, "exploits": [], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-105759", "title": "vLLM: Unbounded Prometheus label cardinality from attacker-controlled HTTP method tokens in the vLLM Rust frontend metrics middleware (unauthenticated denial of service)", "description": "vLLM is an inference and serving engine for large language models. Prior to 0.30.0, the Rust frontend's track_http_metrics middleware records the raw HTTP method token as a Prometheus label for requests reaching registered routes. An unauthenticated attacker can send unique arbitrary method tokens to unguarded routes such as /tokenize, causing Prometheus's Family::get_or_create function to permanently create counter and histogram label sets. Those label sets increase process memory usage and enlarge the /metrics response until the service or monitoring path is exhausted. This issue is fixed in version 0.30.0.", "references": ["https://github.com/vllm-project/vllm/security/advisories/GHSA-5fj9-pfhr-6j48", "https://github.com/vllm-project/vllm/pull/56058", "https://github.com/vllm-project/vllm/commit/3735c2d5f5248259482b9045c34fb7a8a3892352"], "fixes": ["0.30.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}, {"id": "CVE-2026-105758", "noindex": false, "published": "2026-10-05T22:54:53.846Z", "modified": "2026-10-08T02:32:27.776Z", "cvss": 5.3, "cvss_vector": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:L", "severity": "medium", "epss": 0.00303, "kev": false, "exploits": [], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-105758", "title": "vLLM: Qwen2-VL / Qwen3-VL video samplers bound on request-controlled max_frames, which the num_frames ceiling does not reach", "description": "vLLM is an inference and serving engine for large language models. From 0.24.0 until 0.30.0, the Qwen2VLVideoBackend and Qwen3VLVideoBackend classes accept request-level values for the media_io_kwargs.video.max_frames and media_io_kwargs.video.fps fields without enforcing server-side ceilings. An unauthenticated caller can submit these values to the /tokenize endpoint, causing the sampler to decode every frame selected from attacker-controlled video input, consume disproportionate frontend memory, and potentially terminate the API process before scheduling or admission control. The Rust frontend is not affected because it rejects the media_io", "references": ["https://github.com/vllm-project/vllm/security/advisories/GHSA-x6mc-67gf-chw4", "https://github.com/vllm-project/vllm/pull/56729", "https://github.com/vllm-project/vllm/commit/ea723c81c3ea26425cb69503a5d5e90822a04a45"], "fixes": ["0.30.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}, {"id": "CVE-2026-105757", "noindex": false, "published": "2026-10-05T22:52:05.000Z", "modified": "2026-10-06T13:20:25.829Z", "cvss": 6.5, "cvss_vector": "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H", "severity": "medium", "epss": 0.00314, "kev": false, "exploits": [], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-105757", "title": "vLLM: Structured-output request errors escape the request boundary and terminate the shared EngineCore — engine-fatal denial of service (3 sites)", "description": "vLLM is an inference and serving engine for large language models. Prior to 0.30.0, structured-output request failures can escape request-scoped validation and reach the EngineCore fatal-error path. A per-request backend mismatch can re-raise a grammar compilation exception, padding produced by the ngram_gpu speculative-decoding mode can pass a negative token to guidance validation, and the Rust frontend can admit empty structured-output values that the Python frontend rejects, allowing ordinary constrained-generation requests to terminate the shared engine. This issue is fixed in version 0.30.0.", "references": ["https://github.com/vllm-project/vllm/security/advisories/GHSA-85xf-c7hm-whqw", "https://github.com/vllm-project/vllm/pull/51450", "https://github.com/vllm-project/vllm/commit/c55e15a44ec4127832d4a86928a356fdd9e68dbd"], "fixes": ["0.30.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}, {"id": "CVE-2026-105756", "noindex": false, "published": "2026-10-05T22:49:59.642Z", "modified": "2026-10-06T14:42:13.116Z", "cvss": 6.5, "cvss_vector": "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H", "severity": "medium", "epss": 0.00332, "kev": false, "exploits": [], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-105756", "title": "vLLM: Loose `cache_salt` validation lets a single request kill EngineCore on LMCache-MP deployments — uncaught downstream `ValueError` denial of service", "description": "vLLM is an inference and serving engine for large language models. Prior to 0.30.0, OpenAI-compatible request models accept a non-empty cache_salt value without enforcing the character and length restrictions required by the IPCCacheServerKey consumer in LMCache-MP. On deployments using the LMCache-MP connector, a salt that contains a forbidden character or exceeds the permitted length can raise an uncaught ValueError during scheduler cache lookup, causing EngineCore to terminate and denying service to all concurrent users. This issue is fixed in version 0.30.0.", "references": ["https://github.com/vllm-project/vllm/security/advisories/GHSA-2823-qmq8-rwvj", "https://github.com/vllm-project/vllm/pull/51444", "https://github.com/vllm-project/vllm/commit/e962733e08d10f7ca65dac4df99e116460b8b174"], "fixes": ["0.30.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}, {"id": "CVE-2026-105755", "noindex": false, "published": "2026-10-05T22:47:54.854Z", "modified": "2026-10-06T17:55:21.295Z", "cvss": 4.2, "cvss_vector": "CVSS:3.1/AV:N/AC:H/PR:L/UI:N/S:U/C:N/I:L/A:L", "severity": "medium", "epss": 0.00202, "kev": false, "exploits": [], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-105755", "title": "vLLM: Flash late-interaction scoring caches query embeddings under a caller-controlled request id — cross-request integrity break and induced errors on `/score` and `/rerank`", "description": "vLLM is an inference and serving engine for large language models. Prior to 0.30.0, flash late-interaction scoring at the /score and /rerank endpoints derives each worker's query_key value from the caller-controlled X-Request-Id header. A concurrent request that reuses a victim's identifier can overwrite the cached query embedding so the victim's documents are scored against the attacker's query, and shared use counters can also cause a late-interaction cache-miss error. This issue is fixed in version 0.30.0.", "references": ["https://github.com/vllm-project/vllm/security/advisories/GHSA-2phq-3phc-84px", "https://github.com/vllm-project/vllm/pull/51445", "https://github.com/vllm-project/vllm/commit/ee17d0d869203ef9a35ad73358a4987bba14b1fc"], "fixes": ["0.30.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}, {"id": "CVE-2026-105754", "noindex": false, "published": "2026-10-05T22:46:03.163Z", "modified": "2026-10-08T02:31:29.137Z", "cvss": 6.5, "cvss_vector": "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H", "severity": "medium", "epss": 0.00269, "kev": false, "exploits": [], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-105754", "title": "vLLM: Scale-out disaggregated multimodal transport trusts caller-supplied features", "description": "vLLM is an inference and serving engine for large language models. Prior to 0.30.0, the /inference/v1/generate endpoint in the disaggregated scale-out path accepts caller-supplied tensors in the features.kwargs_data field, cache identifiers in the features.mm_hashes field, ranges in the features.mm_placeholders field, and wire-selected multimodal field processors without rebinding them to the active model renderer contract. Forged grid geometry, field types, or non-positive placeholder lengths can terminate the shared EngineCore; when an attacker knows or can induce a victim's content hash, forged cache hashes can poison or retrieve cross-req", "references": ["https://github.com/vllm-project/vllm/security/advisories/GHSA-ph72-cqr5-qpp7", "https://github.com/vllm-project/vllm/pull/51898", "https://github.com/vllm-project/vllm/commit/1970f3ed4be7fa8620e4ddc4a12c36a8384cfc27"], "fixes": ["0.30.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}, {"id": "CVE-2026-105753", "noindex": false, "published": "2026-10-05T22:37:19.633Z", "modified": "2026-10-06T13:23:03.982Z", "cvss": 6.5, "cvss_vector": "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H", "severity": "medium", "epss": 0.00427, "kev": false, "exploits": [], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-105753", "title": "vLLM: Mirrored multimodal IPC caches desync after a rejected request — a later request reusing the same media hash trips a receiver assertion in the engine core", "description": "vLLM is an inference and serving engine for large language models. Prior to 0.28.0, the default mirrored multimodal LRU cache can commit a media hash in the frontend sender cache during multimodal rendering and before engine admission, while the engine receiver cache never receives the payload if that request is rejected. A later request reusing the same media hash causes MultiModalProcessorSenderCache to send no payload and MultiModalReceiverCache to reach an assertion with the message \"Expected a cached item,\" producing a shared-service availability failure. This issue is fixed in version 0.28.0.", "references": ["https://github.com/vllm-project/vllm/security/advisories/GHSA-ph3r-5jfg-f84f", "https://github.com/vllm-project/vllm/pull/46747", "https://github.com/vllm-project/vllm/pull/51897"], "fixes": ["0.28.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}, {"id": "CVE-2026-105752", "noindex": false, "published": "2026-10-05T22:32:59.299Z", "modified": "2026-10-06T14:48:10.692Z", "cvss": 3.1, "cvss_vector": "CVSS:3.1/AV:N/AC:H/PR:L/UI:N/S:U/C:N/I:L/A:N", "severity": "low", "epss": 0.00216, "kev": false, "exploits": ["public-poc"], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-105752", "title": "vLLM: Harmony tool continuations drop `cache_salt` — restoring a cross-tenant prefix-cache membership oracle", "description": "vLLM is an inference and serving engine for large language models. Prior to 0.30.0, Harmony tool continuations submitted through \"POST /v1/responses\" requests rebuild the next-turn engine input without preserving the cache_salt value, placing the continuation prefix in the global unsalted cache namespace even when the caller enabled salting. On deployments with prefix caching enabled, which is the default, an authenticated tenant who can reconstruct a victim's low-entropy post-tool history can submit the same continuation and use the cached_tokens_per_turn count to determine whether the prefix was previously processed, defeating the intended ", "references": ["https://github.com/vllm-project/vllm/security/advisories/GHSA-935w-9g4m-p28p", "https://github.com/vllm-project/vllm/pull/50195", "https://github.com/vllm-project/vllm/pull/51818"], "fixes": ["0.30.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}, {"id": "CVE-2026-103241", "noindex": false, "published": "2026-09-30T16:45:13.581Z", "modified": "2026-10-02T16:23:50.737Z", "cvss": 6.9, "cvss_vector": "CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:L/SC:N/SI:N/SA:N/E:P", "severity": "medium", "epss": 0.00714, "kev": false, "exploits": [], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-103241", "title": "vllm-project vLLM Gemma4UnifiedParser gemma4.rs denial of service", "description": "A flaw has been found in vllm-project vLLM up to 0.26.0. This vulnerability affects unknown code of the file rust/src/parser/src/unified/gemma4.rs of the component Gemma4UnifiedParser. Executing a manipulation can lead to denial of service. The attack may be launched remotely. The exploit has been published and may be used. Upgrading to version 0.29.1rc0 is able to resolve this issue. This patch is called 3439bad37e68ba9755a46f4f6b44a4aeaf1f60a9. Upgrading the affected component is advised.", "references": ["https://vuldb.com/vuln/411965", "https://vuldb.com/vuln/411965/cti", "https://vuldb.com/cve/CVE-2026-103241"], "fixes": ["0.27.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}, {"id": "CVE-2026-100653", "noindex": false, "published": "2026-09-26T13:23:22.605Z", "modified": "2026-10-01T11:06:45.139Z", "cvss": 8.3, "cvss_vector": "CVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:L/VI:H/VA:N/SC:N/SI:N/SA:N", "severity": "high", "epss": 0.0034, "kev": false, "exploits": [], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-100653", "title": "vLLM 0.22.1 before 0.28.0 Incomplete Artifact Pin Propagation", "description": "vLLM is an inference and serving engine for large language models. In versions from 0.22.1 through 0.28.0, the operator-supplied model revision pin (--revision / --code-revision) is not propagated to several Hugging Face artifact loads for the FunAudioChat and Tarsier2 architectures: the WhisperFeatureExtractor and speech_tokenizer PreTrainedTokenizerFast loads in vllm/model_executor/models/funaudiochat.py and the Qwen2VLConfig.from_pretrained call used by Tarsier2ProcessingInfo in vllm/model_executor/models/qwen2_vl.py. As a result, deployments pinned to a reviewed revision still resolve these behavior-affecting processor, tokenizer, and con", "references": ["https://github.com/vllm-project/vllm/security/advisories/GHSA-hhv2-872h-628q", "https://github.com/vllm-project/vllm/commit/d26a28ab033697f55a1414b5b0435de7cd6045b6", "https://www.vulncheck.com/advisories/vllm-0.22.1-before-0.28.0-incomplete-artifact-pin-propagation"], "fixes": ["0.28.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}, {"id": "CVE-2026-100652", "noindex": false, "published": "2026-09-26T13:23:21.865Z", "modified": "2026-09-28T16:42:13.668Z", "cvss": 8.2, "cvss_vector": "CVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N", "severity": "high", "epss": 0.00419, "kev": false, "exploits": [], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-100652", "title": "vLLM 0.22.0 through 0.23.0 Denial of Service via stop_token_ids", "description": "vLLM versions 0.22.0 through 0.23.0 fail to validate stop_token_ids against vocabulary bounds in Rust HTTP and gRPC frontends, allowing out-of-vocabulary token IDs to reach MinTokensLogitsProcessor. Attackers can submit requests with min_tokens greater than zero and out-of-vocabulary stop_token_ids to trigger CUDA tensor indexing failures that leave EngineCore in a fatal state requiring service restart.", "references": ["https://github.com/vllm-project/vllm/security/advisories/GHSA-qff2-492f-9fm4", "https://www.vulncheck.com/advisories/vllm-0.22.0-through-0.23.0-denial-of-service-via-stop-token-ids", "https://nvd.nist.gov/vuln/detail/CVE-2026-100652"], "fixes": ["0.24.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}, {"id": "CVE-2026-100650", "noindex": false, "published": "2026-09-26T13:23:20.418Z", "modified": "2026-10-01T11:06:44.741Z", "cvss": 7.1, "cvss_vector": "CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N", "severity": "high", "epss": 0.00808, "kev": false, "exploits": [], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-100650", "title": "vLLM before 0.29.0 Resource Exhaustion via Unbounded Media Materialization", "description": "vLLM through 0.29.0 fetches and fully materializes remote or inline media before enforcing its documented media controls (the VLLM_MAX_AUDIO_CLIP_FILESIZE_MB compressed-audio size cap, default 25 MB, and the per-modality --limit-mm-per-prompt item limits). Across four ingress paths — the shared media-acquisition layer (HTTPConnection.get_bytes()/async_get_bytes()), the chat completions audio_url/base64 path, the batch speech runner, and the Rust frontend POST /tokenize route — the server reads the entire HTTP response body, base64-decodes the inline payload, or spawns one fetch/decode task per media part, and only then applies the limit (or, ", "references": ["https://github.com/vllm-project/vllm/security/advisories/GHSA-p6g9-7v3x-m8mv", "https://github.com/vllm-project/vllm/commit/752a3a504485790a2e8491cacbb35c137339ad34", "https://www.vulncheck.com/advisories/vllm-before-0.29.0-resource-exhaustion-via-unbounded-media-materialization"], "fixes": ["0.30.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}, {"id": "CVE-2026-94626", "noindex": false, "published": "2026-09-21T22:04:15.352Z", "modified": "2026-09-24T22:54:41.241Z", "cvss": 8.7, "cvss_vector": "CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N", "severity": "high", "epss": 0.00627, "kev": false, "exploits": [], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-94626", "title": "vLLM through 0.29.0 Memory Exhaustion via Unvalidated NIXL tp_size", "description": "vLLM through 0.29.0 fails to validate the tp_size parameter in kv_transfer_params on OpenAI-compatible completion endpoints, allowing attackers to allocate unbounded memory. Attackers can supply arbitrary tp_size values in prefill/decode disaggregated deployments to exhaust memory and trigger kernel OOM-kill of the decode worker process.", "references": ["https://github.com/vllm-project/vllm/pull/51137", "https://github.com/vllm-project/vllm/blob/v0.29.0/vllm/distributed/kv_transfer/kv_connector/utils.py#L569-L573", "https://github.com/vllm-project/vllm/blob/v0.29.0/vllm/distributed/kv_transfer/kv_connector/v1/nixl/metadata.py#L277"], "fixes": ["0.30.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}, {"id": "CVE-2026-94625", "noindex": false, "published": "2026-09-21T22:04:14.206Z", "modified": "2026-09-23T15:09:16.926Z", "cvss": 6.9, "cvss_vector": "CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:L/SC:N/SI:N/SA:N", "severity": "medium", "epss": 0.00521, "kev": false, "exploits": [], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-94625", "title": "vLLM through 0.29.0 Resource Exhaustion via Ownerless Mooncake Transfer Placeholders", "description": "vLLM through 0.29.0 contains a resource exhaustion vulnerability in MooncakeConnector where rejected prefill requests create ownerless transfer placeholders that are never reclaimed. Attackers can send rejected requests to exhaust sender task pools, causing valid requests to be delayed by up to 480 seconds while health checks continue returning success.", "references": ["https://github.com/vllm-project/vllm/pull/51236", "https://github.com/vllm-project/vllm/blob/v0.29.0/vllm/distributed/kv_transfer/kv_connector/v1/mooncake/mooncake_connector.py#L1234-L1242", "https://github.com/vllm-project/vllm"], "fixes": ["0.30.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}, {"id": "CVE-2026-94624", "noindex": false, "published": "2026-09-21T22:04:13.172Z", "modified": "2026-09-22T13:26:57.950Z", "cvss": 8.7, "cvss_vector": "CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N", "severity": "high", "epss": 0.00627, "kev": false, "exploits": [], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-94624", "title": "vLLM through 0.29.0 Denial of Service via Unbounded P2P KV Offloading Sessions", "description": "vLLM through 0.29.0 contains a denial of service vulnerability in P2P KV offloading when OffloadingConnector is configured with TieringOffloadingSpec and a peer-to-peer secondary tier. Attackers can supply arbitrary remote host and port values in kv_transfer_params to create unreachable peer sessions that retain ZeroMQ sockets until the context quota is exhausted, causing an uncaught ZMQError that crashes EngineCore and stops all inference.", "references": ["https://github.com/vllm-project/vllm/pull/51504", "https://github.com/vllm-project/vllm/blob/v0.29.0/vllm/v1/kv_offload/tiering/p2p/control/zmq.py#L240-L260", "https://github.com/vllm-project/vllm/blob/v0.29.0/vllm/v1/kv_offload/tiering/p2p/manager.py#L667-L677"], "fixes": ["0.30.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}, {"id": "CVE-2026-94623", "noindex": false, "published": "2026-09-21T22:04:12.106Z", "modified": "2026-09-22T17:44:56.345Z", "cvss": 8.7, "cvss_vector": "CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N", "severity": "high", "epss": 0.00627, "kev": false, "exploits": [], "stacks": [{"name": "python", "path": "vulns/stack/python", "slug": "python"}], "path": "vulns/cve-2026-94623", "title": "vLLM through 0.29.0 Denial of Service via NIXL Multi-Prompt Assertion Failure", "description": "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 deployments. Attackers can trigger an assertion failure in NixlBaseConnectorWorker._apply_prefix_caching by submitting completion requests with multiple prompts of varying lengths, causing the decode worker to terminate and become unavailable until restarted.", "references": ["https://github.com/vllm-project/vllm/pull/51505", "https://github.com/vllm-project/vllm/blob/v0.29.0/vllm/distributed/kv_transfer/kv_connector/v1/nixl/base_worker.py#L2642-L2646", "https://github.com/vllm-project/vllm"], "fixes": ["0.30.0"], "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}]}], "total": 98, "unknown_count": 0, "release_unconfirmed": false, "page": 1, "pages": 5, "filtered": false, "noindex": false, "filters": {}, "page_size": 20, "products": [{"name": "vllm", "path": "vulns/pypi/vllm", "ecosystem": "PyPI"}], "stacks": [{"name": "system", "path": "vulns/stack/system", "slug": "system"}, {"name": "web", "path": "vulns/stack/web", "slug": "web"}, {"name": "php", "path": "vulns/stack/php", "slug": "php"}, {"name": "python", "path": "vulns/stack/python", "slug": "python"}, {"name": "node", "path": "vulns/stack/node", "slug": "node"}, {"name": "database", "path": "vulns/stack/database", "slug": "database"}, {"name": "containers", "path": "vulns/stack/containers", "slug": "containers"}, {"name": "network", "path": "vulns/stack/network", "slug": "network"}, {"name": "applications", "path": "vulns/stack/applications", "slug": "applications"}], "versions": ["0.31.0", "0.30.0", "0.29.0", "0.28.0", "0.27.1", "0.27.0", "0.26.0", "0.25.1", "0.25.0", "0.24.0", "0.23.0", "0.22.1"], "stats": {"recent": 16, "kev": 0}, "lastmod": "2026-10-21T15:50:53.921Z", "early_advisories": [], "license": "Vous pouvez librement utiliser la sélection de PatchFirst et nos ajouts (priorités, alertes précoces, textes, synthèses et agrégats) pour vos propres besoins, y compris les activités internes de votre entreprise. La revente, la distribution commerciale et la mise en miroir massive de cette sélection ou de ces ajouts nécessitent un accord écrit avec l’opérateur. Lors de toute utilisation ou publication de la sélection ou de nos ajouts, citez PatchFirst (https://patchfirst.dev/vulns/) avec ce lien vers la page du flux de votre région. Les enregistrements originaux des sources ouvertes restent soumis à leurs licences et conditions respectives. Ces conditions ne limitent pas vos droits sur les données originales, que vous pouvez obtenir directement auprès de ces sources.", "terms_url": "https://patchfirst.dev/fr/vulns/#feed-terms", "attribution": {"name": "PatchFirst", "url": "https://patchfirst.dev/vulns/"}, "source_changes": "PatchFirst sélectionne, normalise et combine les enregistrements ; des synthèses et des traductions peuvent être ajoutées. Les actualités sont accompagnées d’un lien vers la source originale dans chaque entrée.", "sources": [{"id": "cve", "name": "CVE List / CVE.org", "url": "https://www.cve.org/", "license": "CVE Terms of Use", "terms_url": "https://www.cve.org/Legal/TermsOfUse", "notice": "Copyright © 1999–2021, The MITRE Corporation. LICENSE Submissions : For all materials you submit to the Common Vulnerabilities and Exposures (CVE™), you hereby grant to The MITRE Corporation (MITRE) and all CVE Numbering Authorities (CNAs) a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable copyright license to reproduce, prepare derivative works of, publicly display, publicly perform, sublicense, and distribute such materials and derivative works. Unless required by applicable law or agreed to in writing, you provide such materials on an \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied, including, without limitation, any warranties or conditions of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A PARTICULAR PURPOSE. CVE Usage : MITRE hereby grants you a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable copyright license to reproduce, prepare derivative works of, publicly display, publicly perform, sublicense, and distribute Common Vulnerabilities and Exposures (CVE™). Any copy you make for such purposes is authorized provided that you reproduce MITRE's copyright designation and this license in any such copy. DISCLAIMERS ALL DOCUMENTS AND THE INFORMATION CONTAINED THEREIN PROVIDED BY MITRE ARE PROVIDED ON AN \"AS IS\" BASIS AND THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE MITRE CORPORATION, ITS BOARD OF TRUSTEES, OFFICERS, AGENTS, AND EMPLOYEES, DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION THEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE."}, {"id": "osv", "name": "OSV", "url": "https://osv.dev/", "terms_url": "https://google.github.io/osv.dev/data/", "terms": "Les licences des bases d’origine varient ; voir le catalogue des sources."}, {"id": "kev", "name": "CISA KEV", "url": "https://www.cisa.gov/known-exploited-vulnerabilities-catalog", "license": "CC0-1.0", "terms_url": "https://github.com/cisagov/kev-data/blob/develop/LICENSE"}, {"id": "epss", "name": "FIRST EPSS", "url": "https://www.first.org/epss/", "terms_url": "https://www.first.org/epss/faq", "terms": "Évaluations publiques gratuites ; mention de la source demandée (FIRST EPSS, évaluations produites par Empirical Security)."}, {"id": "github-advisory", "name": "GitHub Advisory Database", "url": "https://github.com/advisories", "license": "CC-BY-4.0", "terms_url": "https://github.com/github/advisory-database/blob/main/LICENSE.md"}, {"id": "exploitdb", "name": "Exploit-DB", "url": "https://www.exploit-db.com/", "license": "GPL-2.0", "terms_url": "https://gitlab.com/exploit-database/exploitdb/-/blob/main/LICENSE.md"}, {"id": "metasploit", "name": "Metasploit", "url": "https://github.com/rapid7/metasploit-framework", "license": "BSD-3-Clause", "terms_url": "https://github.com/rapid7/metasploit-framework/blob/master/LICENSE", "notice": "Copyright (c) 2006-2026, Rapid7, Inc.\nLicense: BSD-3-clause\n Redistribution and use in source and binary forms, with or without modification,\n are permitted provided that the following conditions are met:\n .\n * Redistributions of source code must retain the above copyright notice,\n this list of conditions and the following disclaimer.\n .\n * Redistributions in binary form must reproduce the above copyright notice,\n this list of conditions and the following disclaimer in the documentation\n and/or other materials provided with the distribution.\n .\n * Neither the name of Rapid7, Inc. nor the names of its contributors\n may be used to endorse or promote products derived from this software\n without specific prior written permission.\n .\n THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\" AND\n ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED\n WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE\n DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR\n ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES\n (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;\n LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON\n ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT\n (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS\n SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE."}, {"id": "nuclei", "name": "Nuclei Templates", "url": "https://github.com/projectdiscovery/nuclei-templates", "license": "MIT", "terms_url": "https://github.com/projectdiscovery/nuclei-templates/blob/main/LICENSE.md", "notice": "MIT License\n\nCopyright (c) 2025 ProjectDiscovery, Inc.\n\nPermission is hereby granted, free of charge, to any person obtaining a copy\nof this software and associated documentation files (the \"Software\"), to deal\nin the Software without restriction, including without limitation the rights\nto use, copy, modify, merge, publish, distribute, sublicense, and/or sell\ncopies of the Software, and to permit persons to whom the Software is\nfurnished to do so, subject to the following conditions:\n\nThe above copyright notice and this permission notice shall be included in all\ncopies or substantial portions of the Software.\n\nTHE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\nIMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\nFITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\nAUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\nLIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\nOUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE\nSOFTWARE."}, {"id": "linux-kernel", "name": "Linux kernel releases", "url": "https://www.kernel.org/", "license": "CC-BY-SA-4.0", "terms_url": "https://www.kernel.org/legal.html", "notice": "Copyright (C) 1997-2014 by The Linux Kernel Organization, Inc."}]}