Android traffic malware analysis and detection using ensemble classifier

Abstract

This paper introduces the Systematic mAlware detection in android (STAR) technique designed to enhance accuracy in identifying and classifying Android malware, addressing significant concerns regarding device security and data privacy. The STAR method involves comprehensive data collection from diverse datasets, rigorous preprocessing for data quality improvement, and feature extraction using Principal Component Analysis (PCA). Butterfly optimization ensures selection of pertinent features, while ensemble classifiers including Bagging, AdaBoost, and LogitBoost are employed for robust model creation. Final classification is achieved via majority voting. Experimental validation demonstrates that STAR outperforms existing techniques such as ERBE, De-LADY, and MSFDROID, achieving detection rates 4.34 %, 1.41 %, and 2.52 % higher respectively. This innovative approach underscores its potential in mitigating the evolving threat landscape of Android malware, offering a promising avenue for enhancing mobile app security.

Keywords

• Malware detection
• Machine learning
• Malware variants
• Malware Classifications