CVE-2026-43622
llama.cpp builds b1886 through b7445 contain a double free vulnerability in the LLaMA-Android JNI wrapper where new_1batch() allocates memory using malloc() while free_1batch() deallocates it using the C++ delete operator, causing heap metadata corruption. Attackers can trigger this memory management mismatch to cause denial of service through process crashes or potentially achieve arbitrary code execution depending on allocator state.
Scoring
- Severity
- HIGH
- CVSS base score
- 8.5
- CVSS vector
- CVSS:4.0/AV:L/AC:L/AT:N/PR:N/UI:P/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N
- EPSS probability
- 0.16%
- CWE
- CWE-762, CWE-415, CWE-762, CWE-415
- Published
- 2026-08-06
- Last modified
- 2026-08-14
Affected products
- ggml-org llama.cpp
- ggml-org llama.cpp
- ggml-org llama.cpp
Weakness type
Related vulnerabilities
- CVE-2025-48431 — Apache Thrift: Specially crafted input can crash a c_glib Thrift server with invalid pointer error.
- CVE-2025-11015 — OGRECave Ogre OgreSTBICodec.cpp encode mismatched memory management routines
- CVE-2025-48755 — In the spiral-rs crate 0.2.0 for Rust, allocation can be attempted for a ZST (zero-sized type).
- CVE-2025-47737 — lib.rs in the trailer crate through 0.1.2 for Rust mishandles allocating with a size of zero.
- CVE-2025-20189 — A vulnerability in the Cisco Express Forwarding functionality of Cisco IOS XE Software for Cisco...
- CVE-2024-4853 — Mismatched Memory Management Routines in editcap
- CVE-2024-2955 — Mismatched Memory Management Routines in Wireshark
- CVE-2023-41056 — Redis vulnerable to integer overflow in certain payloads