Machine Learning in Windows Malware Detection

Abstract

Threats to windows device security are posed by the ever-growing number of malware varieties that can inflict serious harm and spread swiftly, such as viruses, worms, Trojans, and ransomware. These harmful apps have the potential to steal user information, interfere with system operations, and cause losses in money. Against new and polymorphic variants, current malware analysis approaches frequently fall behind in keeping up with the threats that are emerging since they mostly rely on antiquated signature-based techniques that have limited efficacy. Given that the number of malware cases reported each year is increasing substantially, users are more vulnerable as a result of this gap in security. This paper is concerned with a powerful machine learning model made especially for the detection of malware inside Portable Executable (PE) files connected to Windows apps, in order to solve this pressing problem. In our approach, we had 7 classes in our data which motivated us to use tree-based machine learning algorithms. Our proposed models were Random Forest which got 87.3. We were able to achieve higher accuracies than existing models with less computational power and in a more efficient manner.