Malware Detection in PDF and PE Files Using Machine Learning and Feature Selection

Malware detection is a crucial task in cybersecurity. Due to the dynamic nature of malware and the presence of new variants, signature-based malware detection solutions must be complemented by AI and machine learning-based solutions. Malware can hide in different file formats, and use obfuscation techniques to bypass detection. This paper contributes to the body of research by investigating the use of machine learning algorithms and feature selection for the detection of malware in Portable Executable (PE) and Portable Document Format (PDF) files. Different machine learning (ML) algorithms were used. The results from the experiments showed that RF outperformed other algorithms in terms of accuracy, F1 and AUC. Moreover, feature selection can decrease significantly the model building time while maintaining high prediction performance.