An Improved Deep Dense CNN-LSTM Based Malware Identification and Classification

Abstract

Malware identification and classification are receiving greater attention these days as a result of the growing number of attacks on financial and industrial networks. Malware categorization is very tough owing to the exponential increase in the quantity and variety of dangerous files. To have a strong malware defence and post-attack recovery mechanism in place, hostile files must be classified based on their goal, activity, and danger. Malware categorization is an undivided issue that is theoretically NP hard because to the NP hardness of the halting problem. However, as malware has gotten more sophisticated and complicated, traditional tactics have proven more useless. In this paper, we introduced a unique malware classification approach based on convolutional neural networks. CNN had a slightly higher accuracy of 89.7 percent, and when these two were combined (CNN+LSTM), we achieved an accuracy of 92.01 percent.