Automated system for numerical similarity evaluation of Android applications

Abstract

This work is devoted to the design and development of a system for automating the numerical assessment of the similarity of Android applications. The problem of similarity assessment is reduced to the problem of similarity assessment of sets of control flow graphs built on the basis of code from classes.dex files. The similarity value is calculated based on the similarity matrix. Control flow graphs are compared using the graph edit distance and Levenshtein distance algorithms. The paper formulated similarity criteria for Android applications, investigated the forms of their representation, types of models, types of methods for building models, types of assessment of similarity of programs and existing solutions. A prototype of the system and auxiliary tools have been developed, the program has been optimised using parallel programming tools. Experiments were carried out and a conclusion was made about the ability of the program to identify similarities between Android applications.