Dynamic malware detection and phylogeny analysis using process mining

In the last years, mobile phones have become essential communication and productivity tools used daily to access business services and exchange sensitive data. Consequently, they also have become one of the biggest targets of malware attacks. New malware is created everyday, most of which is generated as variants of existing malware by reusing its malicious code. This paper proposes an approach for malware detection and phylogeny studying based on dynamic analysis using process mining. The approach exploits process mining techniques to identify relationships and recurring execution patterns in the system call traces gathered from a mobile application in order to characterize its behavior. The recovered characterization is expressed in terms of a set of declarative constraints between system calls and represents a sort of run-time fingerprint of the application. The comparison between the so defined fingerprint of a given application with those of known malware is used to verify: (1) if the application is malware or trusted, (2) in case of malware, which family it belongs to, and (3) how it differs from other known variants of the same malware family. An empirical study conducted on a dataset of 1200 trusted and malicious applications across ten malware families has shown that the approach exhibits a very good discrimination ability that can be exploited for malware detection and malware evolution studying. Moreover, the study has also shown that the approach is robust to code obfuscation techniques increasingly being used by nowadays malware.