PatchSiren

PatchSiren cyber security CVE debrief

CVE-2026-43630 ggml-org CVE debrief

AI-assisted PatchSiren debrief based on the supplied source corpus. The CVE record was published on 2026-08-06T22:17:06.060Z and has not been modified since then. The NVD entry is currently Received. Users of llama.cpp builds b5702 through b7653, particularly those with write access to the slot save directory, should be aware of this vulnerability and take steps to mitigate it. This includes reviewing and applying patches, restricting write access to the slot save directory, and monitoring server logs for potential exploitation attempts. Additionally, operators, platform administrators, and security teams should assess their exposure and implement compensating controls if necessary. The vulnerability allows attackers with write access to the slot save directory to read memory past the end of the allocated cells array by crafting a malicious slot file with an oversized seq_id value.

Vendor
ggml-org
Product
llama.cpp
CVSS
MEDIUM 6.3
CISA KEV
Not listed in stored evidence
Original CVE published
2026-08-06
Original CVE updated
2026-08-07
Advisory published
2026-08-06
Advisory updated
2026-08-07

Who should care

Users of llama.cpp builds b5702 through b7653, particularly those with write access to the slot save directory, should be aware of this vulnerability and take steps to mitigate it. This includes reviewing and applying patches, restricting write access to the slot save directory, and monitoring server logs for potential exploitation attempts. Additionally, operators, platform administrators, and security teams should assess their exposure and implement compensating controls if necessary.

Technical summary

The vulnerability is caused by an out-of-bounds read in the recurrent memory state restore path of llama.cpp builds b5702 through b7653. Attackers with write access to the slot save directory can craft a malicious slot file with an oversized seq_id value to trigger an out-of-bounds read that leaks heap data including pointer values into server logs. This vulnerability has a CVSS score of 6.3 and is classified as MEDIUM severity. The out-of-bounds read can be used to defeat ASLR protections and facilitate further exploitation.

Defensive priority

Medium-priority defensive actions are recommended due to the CVSS score of 6.3 and the potential for attackers to read memory past the end of the allocated cells array.

Recommended defensive actions

  • Review and apply patches from https://github.com/Vladimir-tokarev-cyera/llama-cpp-security-patches
  • Review and apply patches from https://github.com/ggml-org/llama.cpp/commit/f93c09e2673090b59802f2552848a1cac70598b0
  • Restrict write access to the slot save directory
  • Monitor server logs for potential exploitation attempts
  • Consider compensating controls such as ASLR protections

Evidence notes

The evidence from the NVD and source item indicates a vulnerability in llama.cpp builds b5702 through b7653. The vulnerability allows attackers with write access to the slot save directory to read memory past the end of the allocated cells array by crafting a malicious slot file with an oversized seq_id value. To verify and mitigate this vulnerability, defenders should review the official advisory, assess their exposure, apply patches or mitigations, and monitor for potential exploitation attempts.

Official resources

AI-assisted PatchSiren debrief based on the supplied source corpus. The CVE record was published on 2026-08-06T22:17:06.060Z and has not been modified since then.