An Adaptive Dataset for the Evaluation of Android Malware Detection Techniques

Abstract

Android is currently the leading mobile operating system in the world. The huge number of Android devices attracts developers to create applications for it. However, it also attracts attackers that collect sensitive data or make money. This problem has led many researchers to propose malware detection systems and custom versions of Android that can help users against malicious activities. Evaluating these systems is a crucial part of malware prevention research. However, recent datasets that cover different kinds of benign and malicious applications to evaluate the malware detection techniques are often not available. With thousands of newly released applications every day and different new malicious activities discovered, it is difficult to keep malicious application datasets up to date. This paper introduces a recent and adaptive dataset that includes 5,000 applications from different malware categories that can be used by the research community. The applications are selected from more than 5 million applications. To show how the dataset can be used, we deploy a popular malware analysis platform and generate detailed reports on all the applications in an automated way. We also provide the steps to update the dataset and perform the analysis automatically on the updated set of samples. We believe that the adaptiveness of the dataset and the automatic analysis process will help researchers save time in preparing their datasets and focus more on the detection techniques.