Android Malware Detection

Abstract

Machine learning based detection system of Android malware and analysis of features which are static. The deigned system extracts non executable features such as permissions, intents, activities, and API calls from Android APK files. We analyze non-executable APK features and classify them using a Random Forest model deployed on a Flask server. Our Kotlin application interacts with this backend via REST APIs for real-time prediction. The Android application developed in Kotlin communicates with the backend using REST APIs to perform malware prediction in real time. Training of the model was done using datasets from Debrin, AndroZoo, and custom APK collections, accuracy of 95 Percentage, precision of 94 percentages, and recall of 93 percentages was achieved. The proposed solution provides an effective, scalable, and lightweight detection of malware approach, enhancing mobile security without relying on dynamic or runtime analysis.