Robust Detection of Evasive Fileless Powershell Malware: A Machine Learning Approach

Abstract

In the growing age of cybersecurity, the most obnoxious attack type is PowerShell-based fileless attacks. PowerShell provides the most favored environment to perform advanced tasks. This feature leads to its misuse, especially in the case of fileless attacks. The traditional methods uses signature based detection, are not able to identify the malware. Modern-day scripts are complex and obfuscated, which avoids detection. This paper proposed a machine learning (ML)-based model for malicious sample detection using feature analysis. It efficiently differentiates the benign and malicious samples with a considerable degree of accuracy. To enhance the detection further, the mutual information (MI) technique was applied to retrieve the most efficient and relevant features. This extensive experiment evaluation demonstrated that the proposed ML-based model achieved improved accuracy of 97.64 % and a robust performance.