Reachable Assertion in llama.cpp via gguf_reader

Executive Summary

llama.cpp versions up to b5693 contain a reachable assertion flaw in the gguf_reader::read function. This means an attacker could trigger an assertion failure by providing malformed input, potentially causing the application to crash or behave unexpectedly.

Useful 0 Not Useful 0

Detection Guidance

This vulnerability is specific to the json_schema field in the POST /completions endpoint of llama.cpp versions up to b5693. To detect it, send a deeply nested JSON schema (e.g., 7000+ levels) to the endpoint and monitor for crashes or stack exhaustion. Example curl command: curl -X POST http://localhost:8080/completions -H 'Content-Type: application/json' -d '{"json_schema": {"type": "object", "properties": {"nested": {"type": "object", "properties": {"...": {"type": "object", "properties": {}}}}}}}'.

Useful 0 Not Useful 0

Impact Analysis

This vulnerability could lead to denial-of-service conditions where the affected application crashes. If exploited repeatedly, it may disrupt services relying on llama.cpp for processing.

Useful 0 Not Useful 0

Compliance Impact

This vulnerability (CVE-2026-52131) is a denial-of-service issue caused by uncontrolled recursion in llama.cpp. While it does not directly expose data, a DoS attack could disrupt services handling sensitive data, potentially violating availability requirements in GDPR (Article 32) and HIPAA (Security Rule). However, no specific compliance impact is detailed in the provided context.

Useful 0 Not Useful 0

Mitigation Strategies

Update llama.cpp to a version newer than b5693 to address the Reachable Assertion vulnerability in the gguf_reader::read function.

Useful 0 Not Useful 0