An Interval-Based Model for Detecting Software Defect Using Alias Analysis

Alias analysis is a branch of static program analysis aiming at computing variables which are alias of each other. It is a basis research for many analyses and optimizations in software engineering and compiler construction. Precise modeling of alias analysis is fundamental for software analysis. This paper presents two practical approximation models for representing and computing alias: memory-sensitive model (MSM) and value-sensitive model (VSM). Based on defect-oriented detecting, we present a method to detect software defect using VSM and MSM, which realizes inter-procedure detecting by procedure summary. According to whether type of analysis object coming from defect is value-sensitive or memory-sensitive, we propose two detecting algorithms based on two alias models respectively. One is for memory leak (ML) based on MSM, and the other is for invalid arithmetic operation (IAO) based on VSM. We apply a defect testing system (DTS) to detect six C++ open source projects for proving our models effectiveness. Experimental results show that applying our technique to detect IAO and ML defect can improve detecting efficiency, at the same time reduce potential false positives and false negatives.