Evolutionary Optimization Approaches for Android Malware Detection: A Comparative Assessment of Ant Lion and Whale Optimization

Abstract

With the rise of the Android device market, the rise of Android malware has also made the need for an effective, scalable detection mechanism more pressing and more urgent to meet emerging threats. In this paper, we analyze the importance of Android permissions patterns for malware detection and attempt to tackle the problems arising in high-dimensional feature spaces. For the selection of informative permission features, the wrapper-based metaheuristic optimization algorithms Whale Optimization Algorithm (WOA) and Ant Lion Optimization (ALO) are used to reduce the redundancy of features and enhance the relevance. Several machine learning classifiers such as Decision Tree (DT), Random Forest (RF), Gradient Boosting (GB), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Logistic Regression (LR) are used to evaluate the effectiveness of the optimized feature subsets. Experimental studies show that the Whale Optimization Algorithm-based wrapper feature selection method is the best method, with a classification accuracy of 92.14% and a reduced number of features in the dataset CICInvesAndMal2019 (4,115 permission attributes). The results showed that the feature optimization techniques based on the metaheuristic can be used in improving the accuracy and efficiency of the Android malware detection framework.