Feature Selection Using Adaptive Weight Based Grey Wolf Optimization for Malware Detection in Android

Abstract

The malware application is one of the most dangerous threats to various applications, particularly in Android devices. To effectively detect the malware in Android, the existing researchers utilized the Machine Learning (ML) approaches. However, the researchers have not accurately detected the malware due to the malware complexity, continuous changes as well as increased damages caused by attackers. In this paper, an Adaptive Weight based Grey Wolf Optimization (AWGWO) based feature selection is proposed for malware detection in Android. The proposed AWGWO is evaluated by using Drebin and CICInvesAndMal2019 dataset which contains 216 and 428 malware features. Then, the acquired dataset is pre-processed by utilizing the Min-Max normalization technique to remove the computational complexity and achieve minimum error rates. In the classification stage, a voting combination of benign and malignant malware is used while utilizing three ensemble ML approaches. The ensemble approaches involve a Support Vector Machine (SVM), Random Forest (RF) and AdaBoost algorithm. The proposed AWGWO method achieves better results by using evaluation metrics like accuracy, precision, recall, and F1-score of 0.9953, 0.9930, 0.9963, and 0.9725 respectively which is comparatively better than the existing methods named Random Forest-OWL (RF-OWL), SVM, Adjacency Matrix (AdMat) and Effective Feature selection -RF (EF-RF).