Classification of Polymorphic Virus Based on Integrated Features

Abstract

Standard virus classification relies on the use of virus function, which is a small number of bytes written in assembly language. The addressable problem with current malware intrusion detection and prevention system is having difficulties in detecting unknown and multipath polymorphic computer virus solely based on either static or dynamic features. Thus, this paper presents an effective and efficient polymorphic classification technique based on integrated features. The integrated feature is selected based on Information Gain (IG) rank value between static and dynamic features. Then, all datasets are tested on Naí¯ve Bayes and Random Forest classifiers. We extracted 49 features from 700 polymorphic computer virus samples from Netherland Net Lab and VXHeaven, which includes benign and polymorphic virus function. We spilt the dataset based on 60:40 split ratio sizes for training and testing respectively. Our proposed integrated features manage to achieve 98.9% of accuracy value.