Performance Enhancement of SVM-based ML Malware Detection Model Using Data Preprocessing

The exponential proliferation and speedy propa-gation of malware have been a major concern for computer users. Recently, Machine Learning (ML) is exploited as a sus-tainable solution to combat the increasing volume of malware. Despite its huge success in malware detection, the automated malware classification and detection system face challenges in feature representation and selection perspective. In this work, we investigate a support-vector-machine (SVM)-based machine learning (ML) model for malware detection systems. The focus of this study is to enhance the performance of the SVM-ML model by data preprocessing. Four different data pre-proceeding, transformation, outlier identification, filling and smoothing have been deployed on a standard portable executable header (PEH) classification of mal ware (CLaMP) dataset. The preprocessed dataset is used to train and test the ML model exhibiting an improvement of 7.95% and 3.19% of accuracy for the SVM Linear and Polynomial kernels respectively.