Lucene search
+L
GgmlLlama.cppb1886

4 matches found

CVE
CVE
•added 2026/08/06 3:27 p.m.•89 views

CVE-2026-43622

CVE-2026-43622 affects llama.cpp builds b1886–b7445, where the LLaMA-Android JNI wrapper presents a double free vulnerability. The code allocates memory with malloc() in new_1batch() but frees it with the C++ delete operator in free_1batch(), leading to heap metadata corruption. Consequences desc...

8.5CVSS6.3AI score0.00231EPSS
Save
CVE
CVE
•added 2026/08/06 3:37 p.m.•85 views

CVE-2026-70639

CVE-2026-70639 affects llama.cpp builds b1886–b7445, specifically the LLaMA-Android JNI wrapper. The vulnerability arises because bench_1model() does not validate the model context pointer before dereferencing it, causing a null pointer dereference. An attacker can supply a malformed, corrupt, or...

6.8CVSS5.5AI score0.00195EPSS
Save
CVE
CVE
•added 2026/08/06 3:35 p.m.•75 views

CVE-2026-70638

CVE-2026-70638 affects the llama.cpp project, specifically the LLaMA-Android JNI wrapper. Versions b1886–b7445 are vulnerable due to an integer overflow in new_1batch(), which multiplies sizeof(llama_seq_id) by an attacker-controlled n_seq_max without overflow validation, causing the heap allocat...

8.5CVSS6.5AI score0.00231EPSS
SaveWeb
CVE
CVE
•added 2026/08/06 3:38 p.m.•64 views

CVE-2026-70640

The CVE-2026-70640 issue affects llama.cpp builds b1886–b7445, with a race condition in the LLaMA-Android JNI wrapper: bench_1model() and free_1context() lack synchronization, allowing a thread to operate on freed llama_context memory. Attackers could perform heap spray using attacker-controlled ...

7.3CVSS6.1AI score0.0024EPSS
Save