Efficient Classification of Android Malware in the wild using Robust Static Features

The ubiquitous use of Android smartphones continue to threaten the security and privacy of users’ personal information. Its fast adoption rate makes the smartphone an interesting target for malware authors to deploy new attacks and infect millions of devices. Moreover, the growing number and diversity of malicious applications render conventional defenses ineffective. Thus, there is a need to not only better understand the characteristics of malware families but also, to generate features that are robust and efficient for classification over an extended period of time. In this chapter, we propose a machine learning based malware detection and classification methodology, with the use of static analysis as feature extraction method. Our tool, uniPDroid can be used to extract a plethora of informative features from our extensive dataset. We performed a malware family classification and obtained an average classification accuracy of 92%. We also present the empirical results for our cumulative classification which investigates how well features from old malware can contribute to the detection of new variants of both known and unknown malware.