Conference:
EuroCyberSec 2024, 23. October 2024, Krakow, Poland
Authors:
Barakat R, Josten S, Schneider M.
Abstract:
Buffer overflows remain one of the most critical and widespread vulnerabilities in software systems. Traditional fuzzing techniques often lack the precision required to reliably detect buffer overflows. This paper presents BUFFERMONITOR, a novel approach to enhancing buffer overflow detection by integrating a comprehensive buffer monitoring system into fuzzing frameworks. Using the LLVM compiler framework, we instrument the system under test to collect detailed memory access information, including the distance of each access from buffer boundaries. By prioritizing inputs that generate minimal distances to these boundaries, our method significantly improves the likelihood of detecting potential overflows. This approach not only increases the possibility of identifying buffer overflows but can also identify them with greater accuracy than AddressSanitizer. This provides a robust solution for enhancing software security.
