Semantic code clone detection for enterprise applications

Abstract

Enterprise systems are widely adopted across industries as methods of solving complex problems. As software complexity increases, the software's codebase becomes harder to manage and maintenance costs raise significantly. One such source of cost-raising complexity and code bloat is that of code clones. We proposed an approach to identify semantic code clones in enterprise frameworks by using control flow graphs (CFGs) and applying various proprietary similarity functions to compare enterprise targeted metadata for each pair of CFGs. This approach enables us to detect semantic code clones with high accuracy within a time complexity of O(n2) where n is equal to the number of CFGs composed in the enterprise application (usually around hundreds). We demonstrated our solution on a blind study utilizing a production enterprise application.