Classification of Static Analysis-Based Buffer Overflow Detectors

Abstract

Buffer overflow is one of the most dangerous exploitable vulnerabilities in released software or programs. Many approaches are applied to mitigate buffer overflow (BOF) vulnerabilities such as testing and monitoring. However, BOF vulnerabilities are discovered in programs frequently which might be exploited to crash programs and execute arbitrary injected code. Static analysis is a popular approach for detecting BOF vulnerabilities before releasing programs. Many static analysis-based approaches are currently used in practice. However, there is no detailed classification of these approaches to understand their common characteristics, objectives, and limitations. In this paper, we classify static analysis-based BOF vulnerability detection approaches based on six features: inference technique, analysis sensitivity, analysis granularity, soundness, completeness, and language. We then classify static inference techniques into four types: tainted data flow, constraint, annotation, and string pattern matching. Moreover, we compare the approaches in terms of effectiveness, scalability, and required manual effort. The classification will enable researchers to differentiate among existing analysis approaches. We develop some guidelines to help in choosing approaches and building tools suitable for practitioners need.