Integer Overflow Vulnerability in llama.cpp by ggml-org

CVE-2026-116840

What is CVE-2026-116840?

An integer overflow vulnerability exists in the gguf_init_from_reader function of llama.cpp, prior to version b11275. This flaw allows attackers to exploit crafted GGUF files, leading to a bypass of the size verification intended to prevent processing errors. By supplying a specially crafted tensor size that pads to zero, this vulnerability can result in an undersized data blob. Consequently, this may cause out-of-bounds read operations when consumers attempt to access the improperly allocated memory, potentially leading to information leakage or application crashes.

References