Trojan Detection System Using Machine Learning Approach

Abstract

Malware attack cases continue to rise in our current day. The Trojan attack, which may be extremely destructive by unlawfully controlling other users' computers in order to steal their data. As a result, Trojan horse detection is essential to identify the Trojan and limit Trojan attacks. In this study, we proposed a Trojan detection system that employed machine learning algorithms to detect Trojan horses within the system. A public dataset of Trojan horses that contain 2001 samples comprises of 1041 Trojan horses and 960 of benign is used to train the machine learning classification. In this paper, the Trojan detection system is trained using four types of classifiers which are Random Forest, J48, Decision Table and Naïve Bayes. WEKA is used for the execution of the classification process and performance analysis. The results indicated that the detection system trained with the Random Forest and Decision Table algorithms obtained the maximum level of accuracy.