Android Malware Detection Using Code Graphs

Abstract

The amount of Android malware is increasing faster every year along with the growing popularity of Android platform. Hence, detection and analysis of Android malware have become a critical topic in the area of computer security. This paper proposes a novel method of detecting Android malware that uses the semantics of the code in the form of code graphs extracted from Android apps. These code graphs are then used for classifying Android apps as benign or malicious by using the Jaccard index of the code graphs as a similarity metric. We have also evaluated code graph of real-world Android apps by using the k-NN classifier with Jaccard distance as the distance metric for classification. The result of our experiment shows that code graph of Android apps can be used effectively to detect Android malware with the k-NN classifier, giving a high accuracy of 98%.