Performance Analysis of Deep Learning Algorithms for Malware Detection

Abstract

Malware attack is one of the most critical issues that electronic device users are facing in recent years. New variants of these malwares continue to grow and the traditional approaches like signature based static analysis, dynamic analysis are not suitable for detecting them. So, keeping this in mind, researchers are taking help of Machine Learning and Deep Learning approaches. These approaches have shown a great level of accuracy for determining existing as well as new malwares. In this study, our aim was to analyse the performance of different yet existing algorithms and do the comparison to find the best algorithm amongst them. We choose CNN and LSTM algorithm, and in the end, we combined them to check whether the accuracy is increased or not. After the groundwork, we got 86.50% accuracy for CNN on large dataset. LSTM showed a slightly high accuracy of 89.7% and when these two were combined (CNN+LSTM), the accuracy which we got was 92.01%.