Uncovering Features in Kindred Programs

The detection of similar code can support many software engineering tasks such as program understanding and API replacement. Many excellent approaches have been proposed to detect programs having similar syntactic features. However, some programs dynamically or statistically close to each other, which we call kindred programs, may be ignored. We believe the detection of kindred programs can enhance or even automate the tasks relevant to program classification. In this proposal, we will discuss our current approaches to mine kindred programs having similar functional features and behavioral features. We will also roadmap our on-going development that integrates program analysis with machine learning models to extract statistical features from codebases.