Machine Learning-Driven Android Malware Detection: A Robust Approach

Android malware has become a rapidly growing problem for security. As such, traditional signature-based methods are no longer sufficient for advanced detection mechanisms. This paper reviews ML methods for Android malware detection and classifies the effectiveness in XGBoost, Naive Bayes, and KNN to select the finest model for classification. XGBoost is superior to both models since it uses gradient boosting to construct an ensemble of decision trees that can capture complex feature interactions and dynamically adjust decision boundaries. XGBoost exhibits better performance, with an accuracy of 96.04% which is 26.94% greater than that of Naïve Bayes and 10.2% greater than KNN. Further, XGBoost shows precision of 97.07% which is 30% greater compared to Naïve Bayes and by 15.19% greater when compared to KNN, hence making it the most scalable, efficient, and reliable classifier in the Android Malware detection. This research indicates the capability of XGBoost as ideal solution for strengthening security in android against threats generated by evolving malware.