Securing Mobile Devices from Malware: A Faceoff Between Federated Learning and Deep Learning Models for Android Malware Classification

Abstract

Amidst the escalating threats of android malware, urgency mounts to detect issues while safeguarding user privacy. Traditional machine learning and deep learning methods, dealt with scalability challenges and privacy compromises, finding a potential remedy in federated learning. This study introduces a groundbreaking federated learning-based methodology and compares federated learning with traditional deep learning techniques for Android malware classification, employing renowned datasets, including Drebin, Malgenome, Tuandromd, and Kronodroid. Shifting gears, a federated learning-based approach for malware classification excels in accuracy, scalability, and privacy preservation. Acknowledging limitations and ethical considerations, the study underscores the need for robust privacy measures and dataset transparency. This study unveils federated learning's prowess in android malware classification, opening doors to privacy-driven applications in diverse domains.