Android Dynamic Malware Analysis
Anunay Anand · International Journal for Research in Applied Science and Engineering Technology · 2025
In today’s mobile computing landscape, Android-based systems are highly prevalent and frequently targeted by malicious applications that exhibit anomalous behavior. Detecting such anomalies in real time is critical for ensuring system stability, user data privacy, and overall device security. This review paper explores the implementation and evaluation of unsupervised machine learning techniques for dynamic malware detection in Android applications. The focus is on models such as Isolation Forest, One-Class SVM, Local Outlier Factor, and Elliptic Envelope, which learn from normal process behavior to identify deviations without requiring labeled data. Among these, Isolation Forest demonstrates superior accuracy and efficiency, achieving up to 99% accuracy in detecting anomalous activity based on real-time process metrics like CPU usage, memory consumption, and disk operations. The system is designed to be lightweight, privacy-preserving, and suitable for deployment on individual devices without the need for external infrastructure. This paper also discusses the limitations of existing methods, presents a comparative analysis of model performance, and outlines potential future enhancements including deep learning integration, hybrid detection strategies, and cloud-based intelligence sharing. The findings support the feasibility and effectiveness of machine learning-driven anomaly detection as a proactive defense mechanism in modern Android environments.