{"id":"CVE-2026-33298","aliases":["DEBIAN-CVE-2026-33298","EUVD-2026-14668","GHSA-96jg-mvhq-q7q7","UBUNTU-CVE-2026-33298","openSUSE-SU-2026:12500-1"],"status":"published","title":"llama.cpp has a Heap Buffer Overflow via Integer Overflow in GGUF Tensor Parsing","description":"llama.cpp is an inference of several LLM models in C/C++. Prior to b7824, an integer overflow vulnerability in the `ggml_nbytes` function allows an attacker to bypass memory validation by crafting a GGUF file with specific tensor dimensions. This causes `ggml_nbytes` to return a significantly smaller size than required (e.g., 4MB instead of Exabytes), leading to a heap-based buffer overflow when the application subsequently processes the tensor. This vulnerability allows potential Remote Code Execution (RCE) via memory corruption. b7824 contains a fix.","severity":{"version":"3.1","score":7.8,"vector":"CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H","level":"high","assessor":"cna","from":"cve"},"epss":{"day":"2026-11-11","score":0.00507,"percentile":0.4373,"model":"v2026.06.15"},"facts":{"in_kev":false,"known_exploited":false,"public_exploit":false,"fix_available":true},"cwes":["CWE-122","CWE-190"],"published":"2026-03-24T00:01:40Z","modified":"2026-11-10T13:15:11Z","fixes":[{"ecosystem":"Debian:14","package":"llama.cpp","fixed_in":["7965+dfsg-1"],"sources":["osv"]}],"affected":[{"source":"osv","ecosystem":"Debian:14","package":"llama.cpp","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"7965+dfsg-1"}]}],"versions":["5151+dfsg-1~exp2","5151+dfsg-1~exp3","5318+dfsg-1","5318+dfsg-2","5713+dfsg-1","5760+dfsg-1","5760+dfsg-2","5760+dfsg-3","5760+dfsg-4","5882+dfsg-1","5882+dfsg-2","5882+dfsg-3","5882+dfsg-3~exp1","5882+dfsg-3~exp2","5882+dfsg-3~exp3","5882+dfsg-4","6641+dfsg-1","6641+dfsg-2","6641+dfsg-3","7593+dfsg-1","7593+dfsg-2","7593+dfsg-3"]},{"source":"osv","ecosystem":"Ubuntu:25.10","package":"llama.cpp","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"}]}],"versions":["5318+dfsg-1","5318+dfsg-2","5713+dfsg-1","5760+dfsg-1","5760+dfsg-3","5760+dfsg-4","5882+dfsg-2"]},{"source":"osv","ecosystem":"Ubuntu:26.10","package":"llama.cpp","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"}]}],"versions":["8681+dfsg-1","9721+dfsg-1ubuntu1","10438+dfsg-1","10438+dfsg-1ubuntu1"]},{"source":"osv","ecosystem":"Ubuntu:26.04:LTS","package":"llama.cpp","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"}]}],"versions":["5882+dfsg-2","8064+dfsg-1ubuntu1","8681+dfsg-1"]},{"source":"osv","ecosystem":"openSUSE:Tumbleweed","package":"llamacpp","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"0.9.0-1.1"}]}],"bundled":3}],"products":[{"vendor":"ggml","product":"llama.cpp"},{"vendor":"ggml-org","product":"llama.cpp"}],"references":[{"url":"https://github.com/ggml-org/llama.cpp/security/advisories/GHSA-96jg-mvhq-q7q7","tags":["x_refsource_CONFIRM","Exploit","Vendor Advisory","advisory","report"]},{"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b7824","tags":["x_refsource_MISC","Release Notes","web","report"]},{"url":"https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/33xxx/CVE-2026-33298.json","tags":["advisory"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-33298","tags":["advisory"]},{"url":"https://security-tracker.debian.org/tracker/CVE-2026-33298","tags":["advisory"]},{"url":"https://ubuntu.com/security/CVE-2026-33298","tags":["report"]},{"url":"https://www.cve.org/CVERecord?id=CVE-2026-33298","tags":["report"]},{"url":"","tags":["advisory"]},{"url":"https://www.suse.com/security/cve/CVE-2026-116829","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2026-116840","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2026-33298","tags":["web"]}],"sources":{"cve":[{"id":"CVE-2026-33298","title":"llama.cpp has a Heap Buffer Overflow via Integer Overflow in GGUF Tensor Parsing","description":"llama.cpp is an inference of several LLM models in C/C++. Prior to b7824, an integer overflow vulnerability in the `ggml_nbytes` function allows an attacker to bypass memory validation by crafting a GGUF file with specific tensor dimensions. This causes `ggml_nbytes` to return a significantly smaller size than required (e.g., 4MB instead of Exabytes), leading to a heap-based buffer overflow when the application subsequently processes the tensor. This vulnerability allows potential Remote Code Execution (RCE) via memory corruption. b7824 contains a fix.","status":"published","published":"2026-03-24T00:01:40.989Z","modified":"2026-03-25T03:55:51.679Z","cvss":[{"version":"3.1","score":7.8,"vector":"CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H","level":"high","assessor":"cna"}],"cwes":["CWE-122","CWE-190"],"updated_at":"2026-11-02T14:20:07.609948Z"}],"euvd":[{"id":"EUVD-2026-14668","description":"llama.cpp is an inference of several LLM models in C/C++. Prior to b7824, an integer overflow vulnerability in the `ggml_nbytes` function allows an attacker to bypass memory validation by crafting a GGUF file with specific tensor dimensions. This causes `ggml_nbytes` to return a significantly smaller size than required (e.g., 4MB instead of Exabytes), leading to a heap-based buffer overflow when the application subsequently processes the tensor. This vulnerability allows potential Remote Code Execution (RCE) via memory corruption. b7824 contains a fix.","status":"published","published":"2026-03-24T00:01:40Z","modified":"2026-03-25T03:55:51Z","cvss":[{"version":"3.1","score":7.8,"vector":"CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H","level":"high"}],"updated_at":"2026-11-02T19:25:29.547775Z"}],"nvd":[{"id":"CVE-2026-33298","description":"llama.cpp is an inference of several LLM models in C/C++. Prior to b7824, an integer overflow vulnerability in the `ggml_nbytes` function allows an attacker to bypass memory validation by crafting a GGUF file with specific tensor dimensions. This causes `ggml_nbytes` to return a significantly smaller size than required (e.g., 4MB instead of Exabytes), leading to a heap-based buffer overflow when the application subsequently processes the tensor. This vulnerability allows potential Remote Code Execution (RCE) via memory corruption. b7824 contains a fix.","status":"published","published":"2026-03-24T01:17:01.87Z","modified":"2026-06-17T10:37:16.58Z","cvss":[{"version":"3.1","score":7.8,"vector":"CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H","level":"high","assessor":"secondary"}],"cwes":["CWE-122","CWE-190"],"updated_at":"2026-11-02T14:12:10.568062Z"}],"osv":[{"id":"CVE-2026-33298","title":"llama.cpp has a Heap Buffer Overflow via Integer Overflow in GGUF Tensor Parsing","description":"llama.cpp is an inference of several LLM models in C/C++. Prior to b7824, an integer overflow vulnerability in the `ggml_nbytes` function allows an attacker to bypass memory validation by crafting a GGUF file with specific tensor dimensions. This causes `ggml_nbytes` to return a significantly smaller size than required (e.g., 4MB instead of Exabytes), leading to a heap-based buffer overflow when the application subsequently processes the tensor. This vulnerability allows potential Remote Code Execution (RCE) via memory corruption. b7824 contains a fix.","status":"published","published":"2026-03-24T00:01:40.989Z","modified":"2026-11-10T13:27:18.753687367Z","cvss":[{"version":"3.1","score":7.8,"vector":"CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H","level":"high"}],"cwes":["CWE-122","CWE-190"],"updated_at":"2026-11-10T14:01:20.80945Z"},{"id":"DEBIAN-CVE-2026-33298","description":"llama.cpp is an inference of several LLM models in C/C++. Prior to b7824, an integer overflow vulnerability in the `ggml_nbytes` function allows an attacker to bypass memory validation by crafting a GGUF file with specific tensor dimensions. This causes `ggml_nbytes` to return a significantly smaller size than required (e.g., 4MB instead of Exabytes), leading to a heap-based buffer overflow when the application subsequently processes the tensor. This vulnerability allows potential Remote Code Execution (RCE) via memory corruption. b7824 contains a fix.","status":"published","published":"2026-03-24T01:17:01.87Z","modified":"2026-10-01T03:02:37.341123257Z","cvss":[{"version":"3.1","score":7.8,"vector":"CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H","level":"high"}],"updated_at":"2026-11-02T15:45:54.587575Z"},{"id":"UBUNTU-CVE-2026-33298","description":"llama.cpp is an inference of several LLM models in C/C++. Prior to b7824, an integer overflow vulnerability in the `ggml_nbytes` function allows an attacker to bypass memory validation by crafting a GGUF file with specific tensor dimensions. This causes `ggml_nbytes` to return a significantly smaller size than required (e.g., 4MB instead of Exabytes), leading to a heap-based buffer overflow when the application subsequently processes the tensor. This vulnerability allows potential Remote Code Execution (RCE) via memory corruption. b7824 contains a fix.","status":"published","published":"2026-03-24T01:17:00Z","modified":"2026-10-17T12:59:59.832824925Z","cvss":[{"version":"3.1","score":7.8,"vector":"CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H","level":"high"}],"updated_at":"2026-11-02T16:56:16.036833Z"},{"id":"openSUSE-SU-2026:12500-1","title":"ggml-devel-0.9.0-1.1 on GA media","description":"These are all security issues fixed in the ggml-devel-0.9.0-1.1 package on the GA media of openSUSE Tumbleweed.","status":"published","published":"2026-11-09T00:00:00Z","modified":"2026-11-10T13:15:11.108826119Z","updated_at":"2026-11-10T13:29:51.103746Z"}]}}