Multiclass android malware detection

Abstract

Abstract Android malware poses a significant threat to mobile device security, necessitating robust and efficient detection systems. This project focuses on developing a multiclass classification frame- work for detecting and categorizing Android malware into specific families. Our approach lever- ages a feature extraction pipeline that analyzes critical application components such as intents, permissions, opcodes, and external libraries. These features provide a comprehensive rep- resentation of app behavior and interactions, enabling precise identification of malicious patterns. The classification model is trained on a labeled dataset of Android applications, encompassing both benign and malicious samples from diverse malware families. Advanced machine learn- ing algorithms are employed to discern patterns unique to each family, facilitating multiclass classification. The system’s performance is evaluated using standard metrics such as accuracy, precision, recall, and F1-score, demonstrating its efficacy in distinguishing between multiple malware families. This work contributes to the Android security ecosystem by providing an interpretable, scalable, and effective solution for malware detection and classification, paving the way for enhanced protection against evolving cyber threats.