CVE-2026-70638

Received Received - Intake

Integer Overflow in LLaMA-Android JNI Wrapper

Publication date: 2026-08-06

Last updated on: 2026-08-07

Assigner: VulnCheck

Description

llama.cpp builds b1886 through b7445 contain an integer overflow vulnerability in the LLaMA-Android JNI wrapper where the new_1batch() 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. Attackers can exploit this by providing a crafted n_seq_max value through a malicious model file or JNI call to trigger heap corruption and achieve denial of service or arbitrary code execution on Android applications using the LLaMA-Android binding.

CVSS Scores

EPSS Scores

Probability:
Percentile:

Meta Information

Published

2026-08-06

Last Modified

2026-08-07

Generated

2026-08-27

AI Q&A

2026-08-07

EPSS Evaluated

2026-08-26

NVD

CVE-2026-70638

Affected Vendors & Products

Vendor	Product	Version / Range
llama	llama_cpp	From b1886 (inc) to b7445 (inc)

Helpful Resources

Exploitability

CWE

KEV

CWE ID	Description
CWE-122	A heap overflow condition is a buffer overflow, where the buffer that can be overwritten is allocated in the heap portion of memory, generally meaning that the buffer was allocated using a routine such as malloc().
CWE-190	The product performs a calculation that can produce an integer overflow or wraparound when the logic assumes that the resulting value will always be larger than the original value. This occurs when an integer value is incremented to a value that is too large to store in the associated representation. When this occurs, the value may become a very small or negative number.

Attack-Flow Graph

Executive Summary

This vulnerability is an integer overflow in the LLaMA-Android JNI wrapper of llama.cpp versions b1886 through b7445. The new_1batch() function multiplies sizeof(llama_seq_id) by an attacker-controlled n_seq_max parameter without checking for overflow. This causes incorrect heap buffer allocation, leading to heap corruption. Attackers can exploit this via malicious model files or JNI calls to cause denial of service or arbitrary code execution on Android apps using the LLaMA-Android binding.

Detection Guidance

This vulnerability is specific to the LLaMA-Android JNI wrapper in llama.cpp builds b1886 through b7445. Detection requires checking the version of llama.cpp used in Android applications. Inspect the application's dependencies or source code for the vulnerable JNI wrapper. No direct network or system commands are provided for detection as this is a code-level issue.

Impact Analysis

If you use an Android application that relies on the LLaMA-Android binding from llama.cpp versions b1886 to b7445, an attacker could exploit this flaw to crash the app or execute arbitrary code on your device. This could lead to data theft, unauthorized access, or complete system compromise depending on the app's permissions.

Compliance Impact

This vulnerability could lead to unauthorized code execution or denial of service, potentially compromising data confidentiality and integrity. For GDPR, this may result in unauthorized access to personal data, violating principles of data protection and user rights. For HIPAA, it could expose protected health information, leading to breaches of confidentiality and compliance failures.

Mitigation Strategies

Update llama.cpp to a version beyond b7445 where the integer overflow issue in new_1batch() is fixed. Avoid using untrusted model files or JNI calls with attacker-controlled n_seq_max values.

Hi! I’m here to help you understand CVE-2026-70638. Ask me anything about the vulnerability, its impact, or mitigation strategies.

0/70