PARUDroid: Validation of Android Malware Detection Dataset

Abstract

Android has gained its popularity due to its open nature and number of free apps in its play store. Till date, Android has captured 87% of the total market share. 2.8 million apps are present in the official market of Android. Android apps depend upon permissions for its proper functioning. This dataset contains distinct 5,60,142 Android apps that belong to thirty different categories. These Android application packages (.apk) is collected from Google-play store and other promised repositories. In this study, we performed a dynamic analysis of these collected .apk packages and extracted features, i.e., PARU (Permissions, API calls, Rating of an app, and Users download the app). As per the knowledge, this is the first dataset that extracted features by using the Android 6.0 (API 23) version as an Android operating system. The paper discusses the potential usefulness of the dataset for future research in the field of cybersecurity. Further, to check the potential of our dataset, in this research paper malware detection model is developed by using five different classification machine-learning algorithms. Experiment result reveals that model developed using Deep Neural Network (DNN) can able to detect 98.8% malware-infected apps. Dataset URL: http://dx.doi.org/10.17632/mg5c8jxbhm.2