Integer Overflow in llama.cpp GGUF File Parsing

Executive Summary

This vulnerability is an integer overflow issue in the llama.cpp library affecting versions b10398 through b11300. It occurs in the gguf_init_from_reader function where tensor element-count validation can be bypassed by supplying shapes whose product wraps to zero. Attackers can craft GGUF files with enormous tensor extents that cause overflows in ggml_nbytes() and strides, leading to out-of-bounds memory access when buffers are sized incorrectly.

Useful 0 Not Useful 0

Detection Guidance

To detect this vulnerability, check the version of llama.cpp installed on your system. Versions b10398 through b11300 are vulnerable. Run: `llama.cpp --version` or inspect the library files for version information. Additionally, monitor for crashes or memory corruption when loading GGUF files, as these may indicate exploitation attempts.

Useful 0 Not Useful 0

Impact Analysis

This vulnerability could allow attackers to execute arbitrary code or cause denial of service by exploiting out-of-bounds memory access. It affects systems using vulnerable versions of llama.cpp to load GGUF model files, potentially compromising confidentiality, integrity, and availability of data processed by the application.

Useful 0 Not Useful 0

Mitigation Strategies

Upgrade to version b11301 or later of llama.cpp. This version includes fixes for integer overflows in tensor validation. If upgrading is not immediately possible, avoid loading untrusted GGUF files and implement strict input validation for tensor dimensions to prevent overflow conditions.

Useful 0 Not Useful 0