Statically detecting likely buffer overflow vulnerabilities

TLDR

Buffer overflow attacks may be today's single most important security threat. The paper proposes a new approach to mitigate buffer overflow vulnerabilities by detecting likely vulnerabilities via source‑code analysis and implements it as an extension to the LCLint annotation‑assisted static checking tool. The method exploits semantic comments and lightweight static analyses, implemented as an extension to LCLint. The tool runs as fast as a compiler, is nearly as easy to use, and successfully detected buffer overflow vulnerabilities in two security‑sensitive programs.

Abstract

Buffer overflow attacks may be today's single most important security threat. This paper presents a new approach to mitigating buffer overflow vulnerabilities by detecting likely vulnerabilities through an analysis of the program source code. Our approach exploits information provided in semantic comments and uses lightweight and efficient static analyses. This paper describes an implementation of our approach that extends the LCLint annotation-assisted static checking tool. Our tool is as fast as a compiler and nearly as easy to use. We present experience using our approach to detect buffer overflow vulnerabilities in two security-sensitive programs.