Applying Dataflow Analysis to Detecting Software Vulnerability

In this paper, we propose a software vulnerability checker which takes rules describing vulnerability patterns and a source program as input and detects locations and paths of the patterns in the program. Simple and flow patterns for vulnerabilities are described as rules in the specification language we designed. The lightweight control and data flow analysis is necessary to detect flow patterns. Newly discovered vulnerability patterns can easily be added to the existing rules. We implement the detector in three parts: a pattern matcher which finds locations of vulnerabilities in source program, a flow graph constructor which extracts the control flow and data flow from the program, and a flow analyzer which finds program's vulnerable execution paths.