{"id":"CVE-2026-5497","aliases":["EUVD-2026-36217","GHSA-wcwg-c5fc-9vrc","PYSEC-2026-2302"],"status":"published","title":"Unbounded Frame Count in video/jpeg Base64 Data URL Processing Leads to OOM DoS in vllm-project/vllm","description":"vLLM versions 0.8.0 and later are vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to unbounded frame count processing in the `VideoMediaIO.load_base64()` method. When processing `video/jpeg` data URLs, the method splits the base64 data string on commas to extract individual JPEG frames without enforcing a frame count limit. An attacker can exploit this by crafting a single API request containing thousands of comma-separated base64-encoded JPEG frames in a data URL, causing the server to decode all frames into memory and crash due to excessive memory consumption. This vulnerability is reachable via the OpenAI-compatible chat completions API and does not require authentication.","severity":{"version":"3.0","score":7.5,"vector":"CVSS:3.0/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H","level":"high","assessor":"cna","from":"cve"},"epss":{"day":"2026-11-10","score":0.00537,"percentile":0.45558,"model":"v2026.06.15"},"facts":{"in_kev":false,"known_exploited":false,"public_exploit":false,"fix_available":true},"cwes":["CWE-400","CWE-770"],"published":"2026-06-11T08:31:18Z","modified":"2026-10-28T02:31:02Z","fixes":[{"ecosystem":"PyPI","package":"vllm","fixed_in":["0.19.0"],"sources":["ghsa","osv"]}],"affected":[{"source":"ghsa","ecosystem":"PyPI","package":"vllm","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0.8.0"},{"fixed":"0.19.0"}]}]},{"source":"osv","ecosystem":"PyPI","package":"vllm","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0.8.0"},{"fixed":"0.19.0"}]}],"versions":["0.10.0","0.10.1","0.10.1.1","0.10.2","0.11.0","0.11.1","0.11.2","0.12.0","0.13.0","0.14.0","0.14.1","0.15.0","0.15.1","0.16.0","0.17.0","0.17.1","0.18.0","0.18.1","0.8.0","0.8.1","0.8.2","0.8.3","0.8.4","0.8.5","0.8.5.post1","0.9.0","0.9.0.1","0.9.1","0.9.2"]},{"source":"osv","ecosystem":"PyPI","package":"vllm","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0.8.0"},{"fixed":"0.19.0"}]}],"versions":["0.10.0","0.10.1","0.10.1.1","0.10.2","0.11.0","0.11.1","0.11.2","0.12.0","0.13.0","0.14.0","0.14.1","0.15.0","0.15.1","0.16.0","0.17.0","0.17.1","0.18.0","0.18.1","0.8.0","0.8.1","0.8.2","0.8.3","0.8.4","0.8.5","0.8.5.post1","0.9.0","0.9.0.1","0.9.1","0.9.2"]}],"products":[{"vendor":"red hat","product":"red hat ai inference server"},{"vendor":"red hat","product":"red hat enterprise linux ai (rhel ai) 3"},{"vendor":"red hat","product":"red hat openshift ai (rhoai)"},{"vendor":"vllm","product":"vllm"},{"vendor":"vllm-project","product":"vllm-project/vllm"}],"references":[{"url":"https://huntr.com/bounties/7bd92629-b396-4449-8f88-6c0092530eb4","tags":["exploit","Exploit","Third Party Advisory","web","evidence"]},{"url":"https://github.com/vllm-project/vllm/commit/58ee61422169ce17e08248f8efa1e9df434fe395","tags":["Patch","fix","web"]},{"url":"https://access.redhat.com/security/cve/CVE-2026-5497","tags":["vdb-entry","x_refsource_REDHAT","advisory","web"]},{"url":"https://bugzilla.redhat.com/show_bug.cgi?id=2487813","tags":["issue-tracking","x_refsource_REDHAT","report","web"]},{"url":"https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-5497.json","tags":["x_sadp-csaf-vex","web"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-5497","tags":["advisory"]},{"url":"https://access.redhat.com/errata/RHSA-2026:33524","tags":["web"]},{"url":"https://access.redhat.com/errata/RHSA-2026:33531","tags":["web"]},{"url":"https://github.com/pypa/advisory-database/tree/main/vulns/vllm/PYSEC-2026-2302.yaml","tags":["web"]},{"url":"https://github.com/advisories/GHSA-wcwg-c5fc-9vrc","tags":["advisory"]},{"url":"https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/5xxx/CVE-2026-5497.json","tags":["advisory"]},{"url":"https://github.com/vllm-project/vllm","tags":["package"]}],"sources":{"cve":[{"id":"CVE-2026-5497","title":"Unbounded Frame Count in video/jpeg Base64 Data URL Processing Leads to OOM DoS in vllm-project/vllm","description":"vLLM versions 0.8.0 and later are vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to unbounded frame count processing in the `VideoMediaIO.load_base64()` method. When processing `video/jpeg` data URLs, the method splits the base64 data string on commas to extract individual JPEG frames without enforcing a frame count limit. An attacker can exploit this by crafting a single API request containing thousands of comma-separated base64-encoded JPEG frames in a data URL, causing the server to decode all frames into memory and crash due to excessive memory consumption. This vulnerability is reachable via the OpenAI-compatible chat completions API and does not require authentication.","status":"published","published":"2026-06-11T08:31:18.953Z","modified":"2026-07-22T12:08:48.968Z","cvss":[{"version":"3.0","score":7.5,"vector":"CVSS:3.0/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H","level":"high","assessor":"cna"},{"version":"3.1","score":7.5,"vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H","level":"high","assessor":"adp"}],"cwes":["CWE-400","CWE-770"],"updated_at":"2026-11-02T13:39:27.849196Z"}],"euvd":[{"id":"EUVD-2026-36217","description":"vLLM versions 0.8.0 and later are vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to unbounded frame count processing in the `VideoMediaIO.load_base64()` method. When processing `video/jpeg` data URLs, the method splits the base64 data string on commas to extract individual JPEG frames without enforcing a frame count limit. An attacker can exploit this by crafting a single API request containing thousands of comma-separated base64-encoded JPEG frames in a data URL, causing the server to decode all frames into memory and crash due to excessive memory consumption. This vulnerability is reachable via the OpenAI-compatible chat completions API and does not require authentication.","status":"published","published":"2026-06-11T08:31:18Z","modified":"2026-07-22T12:08:48Z","cvss":[{"version":"3.1","score":7.5,"vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H","level":"high"}],"updated_at":"2026-11-02T18:19:46.900828Z"}],"ghsa":[{"id":"GHSA-wcwg-c5fc-9vrc","title":"vLLM is vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to unbounded frame count processing in the `VideoMediaIO.load_base64()` method","description":"vLLM versions 0.8.0 and later are vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to unbounded frame count processing in the `VideoMediaIO.load_base64()` method. When processing `video/jpeg` data URLs, the method splits the base64 data string on commas to extract individual JPEG frames without enforcing a frame count limit. An attacker can exploit this by crafting a single API request containing thousands of comma-separated base64-encoded JPEG frames in a data URL, causing the server to decode all frames into memory and crash due to excessive memory consumption. This vulnerability is reachable via the OpenAI-compatible chat completions API and does not require authentication.","status":"published","published":"2026-06-11T12:32:44Z","modified":"2026-08-18T20:16:48Z","cvss":[{"version":"3.0","score":7.5,"vector":"CVSS:3.0/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H","level":"high"}],"cwes":["CWE-400","CWE-770"],"updated_at":"2026-11-02T14:41:03.602876Z"}],"nvd":[{"id":"CVE-2026-5497","description":"vLLM versions 0.8.0 and later are vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to unbounded frame count processing in the `VideoMediaIO.load_base64()` method. When processing `video/jpeg` data URLs, the method splits the base64 data string on commas to extract individual JPEG frames without enforcing a frame count limit. An attacker can exploit this by crafting a single API request containing thousands of comma-separated base64-encoded JPEG frames in a data URL, causing the server to decode all frames into memory and crash due to excessive memory consumption. This vulnerability is reachable via the OpenAI-compatible chat completions API and does not require authentication.","status":"published","published":"2026-06-11T10:16:21.903Z","modified":"2026-10-26T11:10:00.157Z","cvss":[{"version":"3.1","score":7.5,"vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H","level":"high","assessor":"secondary"},{"version":"3.0","score":7.5,"vector":"CVSS:3.0/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H","level":"high","assessor":"secondary"}],"cwes":["CWE-400","CWE-770"],"updated_at":"2026-11-02T14:12:23.485267Z"}],"osv":[{"id":"CVE-2026-5497","title":"Unbounded Frame Count in video/jpeg Base64 Data URL Processing Leads to OOM DoS in vllm-project/vllm","description":"vLLM versions 0.8.0 and later are vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to unbounded frame count processing in the `VideoMediaIO.load_base64()` method. When processing `video/jpeg` data URLs, the method splits the base64 data string on commas to extract individual JPEG frames without enforcing a frame count limit. An attacker can exploit this by crafting a single API request containing thousands of comma-separated base64-encoded JPEG frames in a data URL, causing the server to decode all frames into memory and crash due to excessive memory consumption. This vulnerability is reachable via the OpenAI-compatible chat completions API and does not require authentication.","status":"published","published":"2026-06-11T08:31:18.953Z","modified":"2026-10-28T02:31:02.451115665Z","cvss":[{"version":"3.1","score":7.5,"vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H","level":"high"}],"cwes":["CWE-400"],"updated_at":"2026-11-02T15:45:11.22012Z"},{"id":"GHSA-wcwg-c5fc-9vrc","title":"vLLM is vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to unbounded frame count processing in the `VideoMediaIO.load_base64()` method","description":"vLLM versions 0.8.0 and later are vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to unbounded frame count processing in the `VideoMediaIO.load_base64()` method. When processing `video/jpeg` data URLs, the method splits the base64 data string on commas to extract individual JPEG frames without enforcing a frame count limit. An attacker can exploit this by crafting a single API request containing thousands of comma-separated base64-encoded JPEG frames in a data URL, causing the server to decode all frames into memory and crash due to excessive memory consumption. This vulnerability is reachable via the OpenAI-compatible chat completions API and does not require authentication.","status":"published","published":"2026-06-11T12:32:44Z","modified":"2026-08-18T20:40:58.539253503Z","cvss":[{"version":"3.0","score":7.5,"vector":"CVSS:3.0/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H","level":"high"}],"cwes":["CWE-400","CWE-770"],"updated_at":"2026-11-02T15:46:58.401149Z"},{"id":"PYSEC-2026-2302","description":"vLLM versions 0.8.0 and later are vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to unbounded frame count processing in the `VideoMediaIO.load_base64()` method. When processing `video/jpeg` data URLs, the method splits the base64 data string on commas to extract individual JPEG frames without enforcing a frame count limit. An attacker can exploit this by crafting a single API request containing thousands of comma-separated base64-encoded JPEG frames in a data URL, causing the server to decode all frames into memory and crash due to excessive memory consumption. This vulnerability is reachable via the OpenAI-compatible chat completions API and does not require authentication.","status":"published","published":"2026-06-11T10:16:21.903Z","modified":"2026-08-19T12:15:03.324111672Z","cvss":[{"version":"3.1","score":7.5,"vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H","level":"high"}],"updated_at":"2026-11-02T15:52:55.537808Z"}]}}