{"api_version":"1","generated_at":"2026-09-23T17:26:08+00:00","cve":"CVE-2025-66455","urls":{"html":"https://cve.report/CVE-2025-66455","api":"https://cve.report/api/cve/CVE-2025-66455.json","docs":"https://cve.report/api","cve_org":"https://www.cve.org/CVERecord?id=CVE-2025-66455","nvd":"https://nvd.nist.gov/vuln/detail/CVE-2025-66455"},"summary":{"title":"LMDeploy has Remote Code Execution by Pickle Deserialization via handle_zmq_recv in lmdeploy/lmdeploy/pytorch/disagg/conn/engine_conn.py","description":"LMDeploy is a toolkit for compressing, deploying, and serving large language models. Starting in version 0.9.2 and prior to version 0.16.0, LMDeploy's PyTorch DistServe/PD-disaggregation control plane used `recv_pyobj()` to deserialize messages received through a ZeroMQ PULL socket. PyZMQ implements `recv_pyobj()` using Python pickle deserialization, which can execute arbitrary code while reconstructing an object. The peer address used by the receiver was supplied through the `POST /distserve/p2p_connect` HTTP endpoint. An attacker who could reach an affected DistServe API server could cause the server to connect to an attacker-controlled ZeroMQ endpoint and deserialize a crafted pickle payload. API-key authentication is not enabled unless the operator explicitly configures it. As a result, affected DistServe deployments without API keys allowed unauthenticated remote code execution with the privileges of the LMDeploy serving process. This issue affects the PyTorch backend when PD-disaggregation/DistServe is enabled. Ordinary deployments that do not use the affected disaggregated-serving path do not expose this data flow. The fix was released in LMDeploy 0.16.0. Users who cannot upgrade immediately should prevent untrusted clients from reaching `/distserve/*` endpoints, restrict the DistServe HTTP and ZeroMQ control planes to trusted cluster networks, configure API-key authentication, and block arbitrary outbound ZeroMQ connections from serving nodes. These measures reduce exposure but do not make pickle deserialization safe.","state":"PUBLISHED","assigner":"GitHub_M","published_at":"2026-09-18 18:17:04","updated_at":"2026-09-18 20:17:00"},"problem_types":["CWE-502","CWE-502 CWE-502: Deserialization of Untrusted Data"],"metrics":[{"version":"3.1","source":"security-advisories@github.com","type":"Secondary","score":"9.8","severity":"CRITICAL","vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","data":{"version":"3.1","vectorString":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","baseScore":9.8,"baseSeverity":"CRITICAL","attackVector":"NETWORK","attackComplexity":"LOW","privilegesRequired":"NONE","userInteraction":"NONE","scope":"UNCHANGED","confidentialityImpact":"HIGH","integrityImpact":"HIGH","availabilityImpact":"HIGH"}},{"version":"3.1","source":"CNA","type":"DECLARED","score":"9.8","severity":"CRITICAL","vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","data":{"attackComplexity":"LOW","attackVector":"NETWORK","availabilityImpact":"HIGH","baseScore":9.8,"baseSeverity":"CRITICAL","confidentialityImpact":"HIGH","integrityImpact":"HIGH","privilegesRequired":"NONE","scope":"UNCHANGED","userInteraction":"NONE","vectorString":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","version":"3.1"}}],"references":[{"url":"https://github.com/InternLM/lmdeploy/releases/tag/v0.16.0","name":"https://github.com/InternLM/lmdeploy/releases/tag/v0.16.0","refsource":"security-advisories@github.com","tags":[],"title":"","mime":"","httpstatus":"","archivestatus":"0"},{"url":"https://github.com/InternLM/lmdeploy/commit/f05b4ad8bf2e2d84101a1d63b3c44fadd99223b2","name":"https://github.com/InternLM/lmdeploy/commit/f05b4ad8bf2e2d84101a1d63b3c44fadd99223b2","refsource":"security-advisories@github.com","tags":[],"title":"","mime":"","httpstatus":"","archivestatus":"0"},{"url":"https://github.com/InternLM/lmdeploy/security/advisories/GHSA-2vh9-42vm-xmv2","name":"https://github.com/InternLM/lmdeploy/security/advisories/GHSA-2vh9-42vm-xmv2","refsource":"security-advisories@github.com","tags":[],"title":"","mime":"","httpstatus":"","archivestatus":"0"},{"url":"https://www.cve.org/CVERecord?id=CVE-2025-66455","name":"CVE Program record","refsource":"CVE.ORG","tags":["canonical"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2025-66455","name":"NVD vulnerability detail","refsource":"NVD","tags":["canonical","analysis"]}],"affected":[{"source":"CNA","vendor":"InternLM","product":"lmdeploy","version":"affected >= 0.9.2, < 0.16.0","platforms":[]}],"timeline":[],"solutions":[],"workarounds":[],"exploits":[],"credits":[],"nvd_cpes":[],"vendor_comments":[],"enrichments":{"kev":null,"epss":{"cve_year":"2025","cve_id":"66455","cve":"CVE-2025-66455","epss":"0.007020000","percentile":"0.514080000","score_date":"2026-09-21","updated_at":"2026-09-22 00:03:17"},"legacy_qids":[]},"source_records":{"cve_program":{"containers":{"adp":[{"metrics":[{"other":{"content":{"id":"CVE-2025-66455","options":[{"Exploitation":"none"},{"Automatable":"yes"},{"Technical Impact":"total"}],"role":"CISA Coordinator","timestamp":"2026-09-18T19:54:08.825899Z","version":"2.0.3"},"type":"ssvc"}}],"providerMetadata":{"dateUpdated":"2026-09-18T19:54:19.941Z","orgId":"134c704f-9b21-4f2e-91b3-4a467353bcc0","shortName":"CISA-ADP"},"title":"CISA ADP Vulnrichment"}],"cna":{"affected":[{"product":"lmdeploy","vendor":"InternLM","versions":[{"status":"affected","version":">= 0.9.2, < 0.16.0"}]}],"descriptions":[{"lang":"en","value":"LMDeploy is a toolkit for compressing, deploying, and serving large language models. Starting in version 0.9.2 and prior to version 0.16.0, LMDeploy's PyTorch DistServe/PD-disaggregation control plane used `recv_pyobj()` to deserialize messages received through a ZeroMQ PULL socket. PyZMQ implements `recv_pyobj()` using Python pickle deserialization, which can execute arbitrary code while reconstructing an object. The peer address used by the receiver was supplied through the `POST /distserve/p2p_connect` HTTP endpoint. An attacker who could reach an affected DistServe API server could cause the server to connect to an attacker-controlled ZeroMQ endpoint and deserialize a crafted pickle payload. API-key authentication is not enabled unless the operator explicitly configures it. As a result, affected DistServe deployments without API keys allowed unauthenticated remote code execution with the privileges of the LMDeploy serving process. This issue affects the PyTorch backend when PD-disaggregation/DistServe is enabled. Ordinary deployments that do not use the affected disaggregated-serving path do not expose this data flow. The fix was released in LMDeploy 0.16.0. Users who cannot upgrade immediately should prevent untrusted clients from reaching `/distserve/*` endpoints, restrict the DistServe HTTP and ZeroMQ control planes to trusted cluster networks, configure API-key authentication, and block arbitrary outbound ZeroMQ connections from serving nodes. These measures reduce exposure but do not make pickle deserialization safe."}],"metrics":[{"cvssV3_1":{"attackComplexity":"LOW","attackVector":"NETWORK","availabilityImpact":"HIGH","baseScore":9.8,"baseSeverity":"CRITICAL","confidentialityImpact":"HIGH","integrityImpact":"HIGH","privilegesRequired":"NONE","scope":"UNCHANGED","userInteraction":"NONE","vectorString":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","version":"3.1"}}],"problemTypes":[{"descriptions":[{"cweId":"CWE-502","description":"CWE-502: Deserialization of Untrusted Data","lang":"en","type":"CWE"}]}],"providerMetadata":{"dateUpdated":"2026-09-18T17:03:37.084Z","orgId":"a0819718-46f1-4df5-94e2-005712e83aaa","shortName":"GitHub_M"},"references":[{"name":"https://github.com/InternLM/lmdeploy/security/advisories/GHSA-2vh9-42vm-xmv2","tags":["x_refsource_CONFIRM"],"url":"https://github.com/InternLM/lmdeploy/security/advisories/GHSA-2vh9-42vm-xmv2"},{"name":"https://github.com/InternLM/lmdeploy/commit/f05b4ad8bf2e2d84101a1d63b3c44fadd99223b2","tags":["x_refsource_MISC"],"url":"https://github.com/InternLM/lmdeploy/commit/f05b4ad8bf2e2d84101a1d63b3c44fadd99223b2"},{"name":"https://github.com/InternLM/lmdeploy/releases/tag/v0.16.0","tags":["x_refsource_MISC"],"url":"https://github.com/InternLM/lmdeploy/releases/tag/v0.16.0"}],"source":{"advisory":"GHSA-2vh9-42vm-xmv2","discovery":"UNKNOWN"},"title":"LMDeploy has Remote Code Execution by Pickle Deserialization via handle_zmq_recv in lmdeploy/lmdeploy/pytorch/disagg/conn/engine_conn.py"}},"cveMetadata":{"assignerOrgId":"a0819718-46f1-4df5-94e2-005712e83aaa","assignerShortName":"GitHub_M","cveId":"CVE-2025-66455","datePublished":"2026-09-18T17:03:37.084Z","dateReserved":"2025-12-01T18:44:35.640Z","dateUpdated":"2026-09-18T19:54:19.941Z","state":"PUBLISHED"},"dataType":"CVE_RECORD","dataVersion":"5.2"},"nvd":{"publishedDate":"2026-09-18 18:17:04","lastModifiedDate":"2026-09-18 20:17:00","problem_types":["CWE-502","CWE-502 CWE-502: Deserialization of Untrusted Data"],"metrics":{"cvssMetricV31":[{"source":"security-advisories@github.com","type":"Secondary","cvssData":{"version":"3.1","vectorString":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","baseScore":9.8,"baseSeverity":"CRITICAL","attackVector":"NETWORK","attackComplexity":"LOW","privilegesRequired":"NONE","userInteraction":"NONE","scope":"UNCHANGED","confidentialityImpact":"HIGH","integrityImpact":"HIGH","availabilityImpact":"HIGH"},"exploitabilityScore":3.9,"impactScore":5.9}],"ssvcV203":[{"source":"134c704f-9b21-4f2e-91b3-4a467353bcc0","ssvcData":{"timestamp":"2026-09-18T19:54:08.825899Z","id":"CVE-2025-66455","options":[{"exploitation":"none"},{"automatable":"yes"},{"technicalImpact":"total"}],"role":"CISA Coordinator","version":"2.0.3"}}]},"configurations":[]},"legacy_mitre":{"record":{"CveYear":"2025","CveId":"66455","Ordinal":"1","Title":"LMDeploy has Remote Code Execution by Pickle Deserialization via","CVE":"CVE-2025-66455","Year":"2025"},"notes":[{"CveYear":"2025","CveId":"66455","Ordinal":"1","NoteData":"LMDeploy is a toolkit for compressing, deploying, and serving large language models. Starting in version 0.9.2 and prior to version 0.16.0, LMDeploy's PyTorch DistServe/PD-disaggregation control plane used `recv_pyobj()` to deserialize messages received through a ZeroMQ PULL socket. PyZMQ implements `recv_pyobj()` using Python pickle deserialization, which can execute arbitrary code while reconstructing an object. The peer address used by the receiver was supplied through the `POST /distserve/p2p_connect` HTTP endpoint. An attacker who could reach an affected DistServe API server could cause the server to connect to an attacker-controlled ZeroMQ endpoint and deserialize a crafted pickle payload. API-key authentication is not enabled unless the operator explicitly configures it. As a result, affected DistServe deployments without API keys allowed unauthenticated remote code execution with the privileges of the LMDeploy serving process. This issue affects the PyTorch backend when PD-disaggregation/DistServe is enabled. Ordinary deployments that do not use the affected disaggregated-serving path do not expose this data flow. The fix was released in LMDeploy 0.16.0. Users who cannot upgrade immediately should prevent untrusted clients from reaching `/distserve/*` endpoints, restrict the DistServe HTTP and ZeroMQ control planes to trusted cluster networks, configure API-key authentication, and block arbitrary outbound ZeroMQ connections from serving nodes. These measures reduce exposure but do not make pickle deserialization safe.","Type":"Description","Title":"LMDeploy has Remote Code Execution by Pickle Deserialization via"}]}}}