ENSEMBLE APPROACH WITH LASSO REGULARIZATION ALGORITHM FOR IMPROVED DETECTION OF ANDROID MALWARE

Abstract

Android malware is a trending topic in computing and cyber security science, due to over 60% of people usage of android phones. The volume and sophistication of android malware have increased due to this enormous use and creation of different applications which pose serious risks to the security of mobile devices and the services they support. Therefore, there is a growing interest in using machine learning to enhance android malware detection for industry adoption to ensure application security on android phones. This research proposes an efficient ensemble learning approach of the Multilayer Perceptron Neural Network-Random Forest-Light Gradient Boost android malware detection model with Lasso regularisation algorithm for feature selection. Experiment was conducted on two based datasets, CicandMalInvest2019 of a static analysis and TUANDROMD of a dynamic analysis. The results show the data pre-processed with hyperparameter tuning, got an accuracy of 99.601% and 96.92%, false positive rate of 0.0026 and 0.01124, F1-score of 0.91764 and 0.99441, on the static and dynamic analysis datasets respectively. The proposed model outperformed the benchmarked machine learning based models.