Malware Classification Using Xgboost With Vote Based Backward Feature Elimination Technique

Abstract

Malware is one of the most popular threats today, and it is rapidly becoming a significant threat to Internet security. Malware is computer code written by cyber criminals with the intent of causing extensive harm to data and infrastructure or gaining unauthorized access to a network. There are several methods are employed to detect the malware with signature based and behaviour based techniques. Several machine learning techniques are used for classification of malware files. The traditional techniques are not efficient to detect the malware. To efficiently classify the malware, we proposed the XGB with Vote based Backward Feature Elimination technique (XGB-VBFE) which selects the optimal features to build the model and classifies the files with higher accuracy. The performance of the proposed system is compared with other machine learning algorithms such as SVM and Random Forest and proved to be better in accuracy, precision and recall. The proposed XGB-VBFE classifies the files with the accuracy of 99.50%, precision 0.99 and recall 0.96.