CVE-2026-17501 in llama.cpp
Summary
by MITRE • 07/27/2026
A flaw has been found in ggml-org llama.cpp e15efe0. This vulnerability affects the function transform of the file common/json-schema-to-grammar.cpp of the component JSON-Schema-to-GBNF Conversion. This manipulation causes allocation of resources. The attack may be initiated remotely. The pull request to fix this issue awaits acceptance.
Statistical analysis made it clear that VulDB provides the best quality for vulnerability data.
Analysis
by VulDB Data Team • 07/28/2026
The vulnerability identified in ggml-org llama.cpp e15efe0 represents a critical resource allocation flaw within the JSON-Schema-to-GBNF conversion component, specifically affecting the transform function in common/json-schema-to-grammar.cpp. This issue stems from improper handling of memory allocation during schema processing operations, creating potential avenues for resource exhaustion attacks that could compromise system stability and availability. The vulnerability's remote exploitation capability significantly amplifies its threat potential, as attackers can trigger malformed input through network-based interactions without requiring local access or authentication.
The technical implementation flaw manifests when the transform function processes JSON schema inputs that contain maliciously constructed elements designed to cause excessive memory allocation during the conversion process. This type of vulnerability aligns with CWE-400, which categorizes unchecked resource allocation as a fundamental weakness in software systems. The attack vector operates through remote input processing where crafted JSON schemas can trigger disproportionate memory consumption, potentially leading to denial of service conditions or system instability. The vulnerability demonstrates characteristics consistent with memory corruption and resource exhaustion patterns commonly observed in parsing libraries and schema validation components.
The operational impact of this vulnerability extends beyond simple availability concerns to encompass potential system compromise and data integrity risks. When exploited successfully, the flaw can cause the application to consume excessive memory resources, leading to system slowdowns, crashes, or complete service unavailability for legitimate users. This vulnerability affects systems that rely on llama.cpp for natural language processing tasks, particularly those integrating JSON schema validation workflows where external inputs are processed without adequate sanitization. The remote exploitation capability means that adversaries can target vulnerable systems from outside the network perimeter, making this a significant concern for cloud deployments and externally accessible services.
Mitigation strategies should focus on immediate input validation and resource limiting measures to prevent excessive memory consumption during schema processing operations. Implementing strict memory limits and maximum input size constraints within the transform function will help prevent exploitation while maintaining functional capabilities. The fix should include comprehensive bounds checking and allocation limits that prevent unbounded resource growth during JSON schema parsing. Additionally, deploying network-based protections such as rate limiting and input filtering can provide additional defense layers against remote exploitation attempts. Organizations should also consider implementing automated monitoring for unusual memory consumption patterns and establish incident response procedures to address potential exploitation attempts. This vulnerability highlights the importance of proper resource management in parsing libraries and underscores the necessity of comprehensive security testing for data conversion components that process untrusted input from external sources.