{"id":"CVE-2026-70659","aliases":["EUVD-2026-117546","GHSA-6q7q-2v9c-x8m7"],"status":"published","title":"stable-diffusion.cpp: Heap out-of-bounds write in the imatrix loader (`load_imatrix`) via a duplicate tensor name","description":"stable-diffusion.cpp is a pure C/C++ library for running diffusion model (Stable Diffusion, Flux, Wan, Qwen Image, Z-Image, and more) inference. From master-742-1a13107 until master-763-4fcc6fe, IMatrixCollector::load_imatrix in src/runtime/imatrix.cpp reuses the values and counts vectors for entries with the same tensor name but resizes them only when they are empty. An untrusted .imatrix file can provide a small nval for the first entry and a larger nval for a second entry with the same name, causing the accumulation loop to write attacker-controlled floating-point and integer values beyond both heap buffers. Loading the file through the --imat-in conversion path can corrupt the heap or crash the process before a model is required. This issue is fixed in master-763-4fcc6fe.","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":0.00128,"percentile":0.0247,"model":"v2026.06.15"},"facts":{"in_kev":false,"known_exploited":false,"public_exploit":false,"fix_available":false},"cwes":["CWE-1284","CWE-787"],"published":"2026-11-05T15:26:33Z","modified":"2026-11-05T21:32:25Z","products":[{"vendor":"leejet","product":"stable-diffusion.cpp"}],"references":[{"url":"https://github.com/leejet/stable-diffusion.cpp/security/advisories/GHSA-6q7q-2v9c-x8m7","tags":["x_refsource_CONFIRM","advisory"]},{"url":"https://github.com/leejet/stable-diffusion.cpp/pull/1750","tags":["x_refsource_MISC","fix"]},{"url":"https://github.com/leejet/stable-diffusion.cpp/commit/4fcc6fef10c6f686801c6e257b22cd6a3deb906b","tags":["x_refsource_MISC","fix"]},{"url":"https://github.com/leejet/stable-diffusion.cpp/releases/tag/master-763-4fcc6fe","tags":["x_refsource_MISC","web"]},{"url":"https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/70xxx/CVE-2026-70659.json","tags":["advisory"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-70659","tags":["advisory"]}],"sources":{"cve":[{"id":"CVE-2026-70659","title":"stable-diffusion.cpp: Heap out-of-bounds write in the imatrix loader (`load_imatrix`) via a duplicate tensor name","description":"stable-diffusion.cpp is a pure C/C++ library for running diffusion model (Stable Diffusion, Flux, Wan, Qwen Image, Z-Image, and more) inference. From master-742-1a13107 until master-763-4fcc6fe, IMatrixCollector::load_imatrix in src/runtime/imatrix.cpp reuses the values and counts vectors for entries with the same tensor name but resizes them only when they are empty. An untrusted .imatrix file can provide a small nval for the first entry and a larger nval for a second entry with the same name, causing the accumulation loop to write attacker-controlled floating-point and integer values beyond both heap buffers. Loading the file through the --imat-in conversion path can corrupt the heap or crash the process before a model is required. This issue is fixed in master-763-4fcc6fe.","status":"published","published":"2026-11-05T15:26:33.019Z","modified":"2026-11-09T16:51:47.923Z","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-787","CWE-1284"],"updated_at":"2026-11-09T17:34:05.709049Z"}],"euvd":[{"id":"EUVD-2026-117546","description":"stable-diffusion.cpp is a pure C/C++ library for running diffusion model (Stable Diffusion, Flux, Wan, Qwen Image, Z-Image, and more) inference. From master-742-1a13107 until master-763-4fcc6fe, IMatrixCollector::load_imatrix in src/runtime/imatrix.cpp reuses the values and counts vectors for entries with the same tensor name but resizes them only when they are empty. An untrusted .imatrix file can provide a small nval for the first entry and a larger nval for a second entry with the same name, causing the accumulation loop to write attacker-controlled floating-point and integer values beyond both heap buffers. Loading the file through the --imat-in conversion path can corrupt the heap or crash the process before a model is required. This issue is fixed in master-763-4fcc6fe.","status":"published","published":"2026-11-05T15:26:33Z","modified":"2026-11-09T16:51:47Z","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":"2026-11-09T18:17:40.212453Z"}],"nvd":[{"id":"CVE-2026-70659","description":"stable-diffusion.cpp is a pure C/C++ library for running diffusion model (Stable Diffusion, Flux, Wan, Qwen Image, Z-Image, and more) inference. From master-742-1a13107 until master-763-4fcc6fe, IMatrixCollector::load_imatrix in src/runtime/imatrix.cpp reuses the values and counts vectors for entries with the same tensor name but resizes them only when they are empty. An untrusted .imatrix file can provide a small nval for the first entry and a larger nval for a second entry with the same name, causing the accumulation loop to write attacker-controlled floating-point and integer values beyond both heap buffers. Loading the file through the --imat-in conversion path can corrupt the heap or crash the process before a model is required. This issue is fixed in master-763-4fcc6fe.","status":"published","published":"2026-11-05T16:17:48.047Z","modified":"2026-11-09T17:17:32.323Z","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-787","CWE-1284"],"updated_at":"2026-11-09T20:15:34.274131Z"}],"osv":[{"id":"CVE-2026-70659","title":"stable-diffusion.cpp: Heap out-of-bounds write in the imatrix loader (`load_imatrix`) via a duplicate tensor name","description":"stable-diffusion.cpp is a pure C/C++ library for running diffusion model (Stable Diffusion, Flux, Wan, Qwen Image, Z-Image, and more) inference. From master-742-1a13107 until master-763-4fcc6fe, IMatrixCollector::load_imatrix in src/runtime/imatrix.cpp reuses the values and counts vectors for entries with the same tensor name but resizes them only when they are empty. An untrusted .imatrix file can provide a small nval for the first entry and a larger nval for a second entry with the same name, causing the accumulation loop to write attacker-controlled floating-point and integer values beyond both heap buffers. Loading the file through the --imat-in conversion path can corrupt the heap or crash the process before a model is required. This issue is fixed in master-763-4fcc6fe.","status":"published","published":"2026-11-05T15:26:33.019Z","modified":"2026-11-11T07:24:31.789543298Z","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-1284","CWE-787"],"updated_at":"2026-11-07T03:08:09.473992Z"}]}}