{"id":"CVE-2026-116180","aliases":["EUVD-2026-109441","GHSA-rhcx-5729-88vg","GHSA-v8gv-gxpf-cg62"],"status":"published","title":"vLLM from 0.8.5 before 0.31.0 Cross-Tenant DoS via Uninitialized LogprobsTensors","description":"vLLM from 0.8.5 before 0.31.0 has a denial of service flaw. LogprobsTensors.empty_cpu() in vllm/v1/outputs.py allocates an uninitialized int32 tensor, and the V1 model runner leaves prompt positions reported as computed by a KV connector unwritten. A client sharing the engine can send prompt_logprobs (or echo with logprobs) against a connector-cached prefix, making tokenizer decode raise OverflowError, which crashes EngineCore and drops all co-tenant requests.","severity":{"version":"4.0","score":8.3,"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:H","level":"high","assessor":"cna","from":"cve"},"epss":{"day":"2026-11-12","score":0.00254,"percentile":0.17151,"model":"v2026.06.15"},"facts":{"in_kev":false,"known_exploited":false,"public_exploit":false,"fix_available":false},"cwes":["CWE-908"],"published":"2026-10-25T13:20:19Z","modified":"2026-10-30T02:46:48Z","products":[{"vendor":"vllm","product":"vllm"},{"vendor":"vllm-project","product":"vllm"}],"references":[{"url":"https://github.com/vllm-project/vllm/security/advisories/GHSA-rhcx-5729-88vg","tags":["vendor-advisory","Mitigation","Vendor Advisory","advisory"]},{"url":"https://www.vulncheck.com/advisories/vllm-from-0.8.5-before-0.31.0-cross-tenant-dos-via-uninitialized-logprobstensors","tags":["third-party-advisory","Third Party Advisory","advisory"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-116180","tags":["advisory"]},{"url":"https://github.com/advisories/GHSA-v8gv-gxpf-cg62"},{"url":"https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/116xxx/CVE-2026-116180.json","tags":["advisory"]}],"sources":{"cve":[{"id":"CVE-2026-116180","title":"vLLM from 0.8.5 before 0.31.0 Cross-Tenant DoS via Uninitialized LogprobsTensors","description":"vLLM from 0.8.5 before 0.31.0 has a denial of service flaw. LogprobsTensors.empty_cpu() in vllm/v1/outputs.py allocates an uninitialized int32 tensor, and the V1 model runner leaves prompt positions reported as computed by a KV connector unwritten. A client sharing the engine can send prompt_logprobs (or echo with logprobs) against a connector-cached prefix, making tokenizer decode raise OverflowError, which crashes EngineCore and drops all co-tenant requests.","status":"published","published":"2026-10-25T13:20:19.199Z","modified":"2026-10-26T19:48:35.789Z","cvss":[{"version":"4.0","score":8.3,"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:H","level":"high","assessor":"cna"},{"version":"3.1","score":7.7,"vector":"CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:N/I:N/A:H","level":"high","assessor":"cna"}],"cwes":["CWE-908"],"updated_at":"2026-11-02T14:20:03.827373Z"}],"euvd":[{"id":"EUVD-2026-109441","description":"vLLM from 0.8.5 before 0.31.0 has a denial of service flaw. LogprobsTensors.empty_cpu() in vllm/v1/outputs.py allocates an uninitialized int32 tensor, and the V1 model runner leaves prompt positions reported as computed by a KV connector unwritten. A client sharing the engine can send prompt_logprobs (or echo with logprobs) against a connector-cached prefix, making tokenizer decode raise OverflowError, which crashes EngineCore and drops all co-tenant requests.","status":"published","published":"2026-10-25T13:20:19Z","modified":"2026-10-26T19:48:35Z","cvss":[{"version":"4.0","score":8.3,"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:H","level":"high"}],"updated_at":"2026-11-02T16:59:28.357127Z"}],"ghsa":[{"id":"GHSA-v8gv-gxpf-cg62","title":"vLLM from 0.8.5 before 0.31.0 has a denial of service flaw. LogprobsTensors.empty_cpu() in vllm...","description":"vLLM from 0.8.5 before 0.31.0 has a denial of service flaw. LogprobsTensors.empty_cpu() in vllm/v1/outputs.py allocates an uninitialized int32 tensor, and the V1 model runner leaves prompt positions reported as computed by a KV connector unwritten. A client sharing the engine can send prompt_logprobs (or echo with logprobs) against a connector-cached prefix, making tokenizer decode raise OverflowError, which crashes EngineCore and drops all co-tenant requests.","status":"published","published":"2026-10-25T15:32:20Z","modified":"2026-10-25T15:32:31Z","cvss":[{"version":"4.0","score":8.3,"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:H/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X","level":"high"},{"version":"3.1","score":7.7,"vector":"CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:N/I:N/A:H","level":"high"}],"cwes":["CWE-908"],"updated_at":"2026-11-02T15:14:22.728467Z"}],"nvd":[{"id":"CVE-2026-116180","description":"vLLM from 0.8.5 before 0.31.0 has a denial of service flaw. LogprobsTensors.empty_cpu() in vllm/v1/outputs.py allocates an uninitialized int32 tensor, and the V1 model runner leaves prompt positions reported as computed by a KV connector unwritten. A client sharing the engine can send prompt_logprobs (or echo with logprobs) against a connector-cached prefix, making tokenizer decode raise OverflowError, which crashes EngineCore and drops all co-tenant requests.","status":"published","published":"2026-10-25T14:16:57.76Z","modified":"2026-10-28T15:40:03.157Z","cvss":[{"version":"4.0","score":8.3,"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:H/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X","level":"high","assessor":"secondary"},{"version":"3.1","score":7.7,"vector":"CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:N/I:N/A:H","level":"high","assessor":"secondary"}],"cwes":["CWE-908"],"updated_at":"2026-11-02T14:13:03.126833Z"}],"osv":[{"id":"CVE-2026-116180","title":"vLLM from 0.8.5 before 0.31.0 Cross-Tenant DoS via Uninitialized LogprobsTensors","description":"vLLM from 0.8.5 before 0.31.0 has a denial of service flaw. LogprobsTensors.empty_cpu() in vllm/v1/outputs.py allocates an uninitialized int32 tensor, and the V1 model runner leaves prompt positions reported as computed by a KV connector unwritten. A client sharing the engine can send prompt_logprobs (or echo with logprobs) against a connector-cached prefix, making tokenizer decode raise OverflowError, which crashes EngineCore and drops all co-tenant requests.","status":"published","published":"2026-10-25T13:20:19.199Z","modified":"2026-10-30T02:46:48.311535872Z","cvss":[{"version":"4.0","score":8.3,"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:H","level":"high"}],"cwes":["CWE-908"],"updated_at":"2026-11-02T15:44:56.675758Z"}]}}