These pages are published after PatchSiren validates generated defensive summaries against stored public CVE and source evidence.
A vulnerability was detected in ggml-org llama.cpp up to 0.4.0, impacting the rpc_server::deserialize_tensor function in ggml/src/ggml-rpc/ggml-rpc.cpp of the RPC Server component. This results in a reachable assertion through manipulation of the argument ne. The attack can be carried out remotely. Defenders should assess exposure and verify configurations. The CVE record and associated details provide li [truncated]
The CVE record for CVE-2026-52131 was published on 2026-09-01T18:17:43.717Z and has not been modified since then. The NVD entry is currently Analyzed. This HIGH severity vulnerability affects llama.cpp versions up to b5693, involving a Reachable Assertion via the gguf_reader::read function. Security teams and developers using llama.cpp, especially those using versions up to b5693, should be aware of this [truncated]
The CVE record for CVE-2026-52130 was published on 2026-09-01T18:17:43.603Z and has not been modified since then. The NVD entry is currently Analyzed. This vulnerability affects llama.cpp, specifically in common/json-schema-to-grammar.cpp, allowing for uncontrolled recursion, which could lead to a denial of service. Security teams and administrators responsible for systems using llama.cpp should be aware [truncated]
AI-assisted PatchSiren debrief based on the supplied source corpus. The CVE record was published on 2026-08-21T17:16:30.810Z and has not been modified since then. The NVD entry is currently Received. CVE-2026-39909 is a use-after-free vulnerability in the RPC server's GRAPH_RECOMPUTE handler of llama.cpp before version b8585. This vulnerability allows unauthenticated remote attackers to achieve arbitrary [truncated]
AI-assisted PatchSiren debrief based on the supplied source corpus. The CVE record was published on 2026-08-06T22:18:28.927Z and has not been modified since then. The NVD entry is currently Awaiting Analysis. This race condition use-after-free vulnerability exists in llama.cpp builds b1886 through b7445, specifically in the LLaMA-Android JNI wrapper affecting functions bench_1model() and free_1context(). [truncated]
The LLaMA-Android JNI wrapper in llama.cpp builds b1886 through b7445 contains an integer overflow vulnerability in the new_1batch() function. This function multiplies sizeof(llama_seq_id) by an attacker-controlled n_seq_max parameter without overflow validation, causing heap buffer allocation to wrap and allocate insufficient memory. This vulnerability can lead to denial of service or arbitrary code exec [truncated]
AI-assisted PatchSiren debrief based on the supplied source corpus. The CVE record was published on 2026-08-06T22:17:06.343Z and has not been modified since then. CVE-2026-43632 is a critical use-after-free vulnerability in llama.cpp, affecting versions b7492 through b9060. The vulnerability is located in the llama-server and impacts six tokenization endpoints: /tokenize, /detokenize, /infill, /apply-temp [truncated]
CVE-2026-43631 is a critical use-after-free vulnerability in llama.cpp builds b7492 through b9060. The vulnerability exists in the vocab pointer of llama-server when the --sleep-idle-seconds feature is enabled, allowing unauthenticated remote attackers to execute arbitrary code. Affected product deployments should be identified and owners assigned for follow-up. Official advisories and CVE records should [truncated]
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 [truncated]
AI-assisted PatchSiren debrief based on the supplied source corpus. The CVE record was published on 2026-08-06T22:17:05.917Z and has not been modified since then. The NVD entry is currently Received. CVE-2026-43629 is a critical heap buffer overflow vulnerability in the KV cache state restore path of llama.cpp builds b4882 through b9058. The state_read_data() function computes write size without overflow [truncated]
AI-assisted PatchSiren debrief based on the supplied source corpus. The CVE record was published on 2026-08-06T22:17:05.777Z and has not been modified since then. The CVE-2026-43628 vulnerability is an integer underflow and out-of-bounds read issue in the DRY sampler of llama.cpp, allowing unauthenticated attackers to trigger a heap buffer underflow by sending a crafted HTTP request with dry_allowed_lengt [truncated]
AI-assisted PatchSiren debrief based on the supplied source corpus. The CVE record for CVE-2026-43627 was published on 2026-08-06T22:17:05.633Z and has not been modified since then. This CVE record indicates an integer overflow vulnerability in llama.cpp builds b1283 through b9058. The vulnerability is in the llama_batch_init() function where unchecked multiplications in malloc() calls can wrap past INT32 [truncated]
AI-assisted PatchSiren debrief based on the supplied source corpus. The CVE record was published on 2026-08-06T16:16:42.883Z and has not been modified since then. The NVD entry is currently Received. CVE-2026-43622 is a double free vulnerability in the LLaMA-Android JNI wrapper of llama.cpp builds b1886 through b7445. Memory allocated using malloc() is deallocated using the C++ delete operator, leading to [truncated]
A vulnerability was found in ggml-org whisper.cpp 95ea8f9b. Affected is the function ggml_ftype_to_ggml_type of the file ggml/src/ggml.c. The manipulation of the argument ftype results in reachable assertion. The attack requires a local approach. This issue has been publicly disclosed and may impact ggml-org whisper.cpp users with local access. The vulnerability class is related to an assertion that can b [truncated]
A flaw has been found in ggml-org llama.cpp e15efe0, affecting the function transform of the file common/json-schema-to-grammar.cpp of the component JSON-Schema-to-GBNF Conversion. This manipulation causes allocation of resources. The attack may be initiated remotely. The CVSS score is 6.9, indicating a medium severity. Users and administrators of ggml-org llama.cpp e15efe0 should review the vulnerability [truncated]
AI-assisted PatchSiren debrief based on the supplied source corpus. The CVE record was published on 2026-07-27T01:16:26.380Z and has not been modified since then. This vulnerability affects ggml-org llama.cpp installations, specifically the function _visit_pattern of the file common/json-schema-to-grammar.cpp, leading to a null pointer dereference. The attack can be launched remotely with a CVSS score of [truncated]