CVE-2026-70640

  1. OpenCVE
  2. Vulnerabilities (CVE)
  3. CVE-2026-70640
llama.cpp builds b1886 through b7445 contain a race condition use-after-free vulnerability in the LLaMA-Android JNI wrapper where bench_1model() and free_1context() lack synchronization, allowing Thread A to operate on freed memory while Thread B concurrently frees the llama_context. Attackers can exploit this by performing heap spray with attacker-controlled data containing a fake vtable to hijack the vtable pointer at offset +0x30, causing llama_batch_allocr::clear() to dereference arbitrary memory and achieve remote code execution.

7.0 /10

CVSS v3 : HIGH

V3 Legend

Vector :

Exploitability : 1.0 / Impact : 5.9

Attack Vector LOCAL

Attack Complexity HIGH

Privileges Required NONE

User Interaction REQUIRED

Confidentiality Impact HIGH

Integrity Impact HIGH

Availability Impact HIGH

Scope UNCHANGED

References
Link Resource
https://github.com/Vladimir-tokarev-cyera/llama-cpp-security-patches Patch Third Party Advisory
https://github.com/ggml-org/llama.cpp/commit/5c0d18881e0e9794c96b2602736b758bac9d9388 Patch
https://github.com/ggml-org/llama.cpp/releases/tag/b7446 Patch Release Notes Vendor Advisory
https://www.vulncheck.com/advisories/llama-cpp-b1886-b7445-race-condition-use-after-free-via-llama-android-cpp Exploit Third Party Advisory
Configurations

Configuration 1 (hide)

cpe:2.3:a:ggml:llama.cpp:*:*:*:*:*:*:*:*
History

23 Oct 2026, 21:10

Type Values Removed Values Added
Summary
  • (es) Las versiones b1886 a b7445 de llama.cpp contienen una vulnerabilidad de uso después de liberación por condición de carrera en el envoltorio JNI de LLaMA-Android donde bench_1model() y free_1context() carecen de sincronización, permitiendo que el Hilo A opere sobre memoria liberada mientras el Hilo B libera concurrentemente el llama_context. Los atacantes pueden explotar esto realizando una pulverización de heap con datos controlados por el atacante que contengan una vtable falsa para secuestrar el puntero de vtable en el desplazamiento +0x30, haciendo que llama_batch_allocr::clear() desreferencie memoria arbitraria y logre la ejecución remota de código.

21 Sep 2026, 18:10

Type Values Removed Values Added
First Time Ggml
Ggml llama.cpp
References () https://github.com/Vladimir-tokarev-cyera/llama-cpp-security-patches - () https://github.com/Vladimir-tokarev-cyera/llama-cpp-security-patches - Patch, Third Party Advisory
References () https://github.com/ggml-org/llama.cpp/commit/5c0d18881e0e9794c96b2602736b758bac9d9388 - () https://github.com/ggml-org/llama.cpp/commit/5c0d18881e0e9794c96b2602736b758bac9d9388 - Patch
References () https://github.com/ggml-org/llama.cpp/releases/tag/b7446 - () https://github.com/ggml-org/llama.cpp/releases/tag/b7446 - Patch, Release Notes, Vendor Advisory
References () https://www.vulncheck.com/advisories/llama-cpp-b1886-b7445-race-condition-use-after-free-via-llama-android-cpp - () https://www.vulncheck.com/advisories/llama-cpp-b1886-b7445-race-condition-use-after-free-via-llama-android-cpp - Exploit, Third Party Advisory
CPE cpe:2.3:a:ggml:llama.cpp:*:*:*:*:*:*:*:*
Information

Published : 2026-08-06 22:18

Updated : 2026-10-29 21:17

NVD link : CVE-2026-70640

Mitre link : CVE-2026-70640

CVE.ORG link : CVE-2026-70640

JSON object : View

Products Affected

ggml

  • llama.cpp
CWE
CWE-362

Concurrent Execution using Shared Resource with Improper Synchronization ('Race Condition')

CWE-476

NULL Pointer Dereference