Malware Detection Using Machine Learning Algorithms for Windows Platform

Abstract

Windows is a popular Graphical User Interface-based Operating System that provides services like storage, run third-party software, play videos, network connection, etc. The purpose of such services can be demolished by targeting the availability of these services. Malware is one of the major security concerns for the Windows platform. Malware is any type of computer software that disturbs the availability of computer services. The traditional detection systems such as the intrusion detection/prevention system, Anti-Virus software cannot detect unseen malware due to the use of signature-based methods. So, there is a need to accurately detect such kind of malware in the Windows environment. In this work, a Machine Learning (ML)-based malware detection system is introduced which extracts features from the Portable Executable file's header to detect whether the executable is clean or malicious. After preprocessing the data, several ML models including Random Forest, Support Vector Machine (SVM), Decision Tree, AdaBoost, Gaussian Naive Bayes (GNB), and Gradient Boosting are applied to cope up with the malware. Moreover, a comparative analysis is conducted among ML models to select the appropriate one for the targeted problem. The experimental results show that the Random Forest outperformed the others with an accuracy level of 99.44\% for the detection of malware. This can be used to develop a desktop application for scanning the malware for the Windows platform with the added ability to customize the scanning process.