Gauss-Mapping Black Widow Optimization With Deep Extreme Learning Machine for Android Malware Classification Model

Abstract

Nowadays, the malware on the Android platform is found to be increasing. With the prevalent use of code obfuscation technology, the precision of antivirus software and classical detection techniques is low. Classical detection techniques of signature matching and manual analysis have exposed issues like low accuracy and slow detection speed. Several authors have overcome the issue of Android malware detection utilizing machine learning (ML) techniques and had more research outcomes. With the growth of deep learning (DL), many researchers started to use DL methods for detecting Android malware. This article introduces a Gauss-Mapping Black Widow Optimization with Deep Learning Enabled Android Malware Classification (GBWODL-AMC) model. The major intention of the GBWODL-AMC technique lies in the automated classification of Android malware. To accomplish this, the GBWODL-AMC technique involves the design of GBWO based feature selection approach to enhance the classification performance. For Android malware classification purposes, the GBWODL-AMC technique employs a deep extreme learning machine (DELM) model and its parameter are optimally selected by the ant lion optimization (ALO) algorithm. The simulation analysis of the GBWODL-AMC technique is tested on CICAndMal2017 dataset. Extensive experimental results signify the better performance of the GBWODL-AMC technique over other malware detectors with maximum accuracy of 98.95%.

Keywords

• Android malware
• machine learning
• cybersecurity
• feature selection
• parameter tuning