A Comprehensive Review of Machine Learning Approaches for Android Malware Detection
Aneesha Davarasan, Joshua Samual, Kulothunkan Palansundram, Aitizaz Ali
Abstract
In today's digital age, smartphones have evolved beyond communication devices, becoming integral to various aspects of daily life. Android, as a leading mobile operating system, dominates the market due to its open-source nature and vast user base. However, this widespread adoption has made it a prime target for increasingly sophisticated malware attacks. Traditional malware detection methods, primarily reliant on signature recognition, have proven insufficient in countering these dynamic threats. This paper provides a detailed review of Android malware detection approaches leveraging machine learning techniques. By examining the underlying Android architecture and security models, we explore static, dynamic, and hybrid analysis methods, highlighting the crucial role of feature selection in improving detection accuracy. Additionally, we address the significant challenges posed by deterioration in detection model performance over time and evasion tactics employed by malware, proposing advanced strategies such as adversarial training and regular model updates to enhance system resilience. This review aims to synthesize current methodologies, offering a critical evaluation and identifying potential avenues for future research to fortify Android malware detection systems.