DL-MD: Leveraging LSTM and GRU for Cutting-Edge Malware Detection

In the field of cyber-security, identifying and combating malware threats continues to pose a significant challenge. While traditional techniques like signature detection and behavioral analysis have shown success, they struggle to keep up with the evolving landscape of malware variations. Additionally, although traditional clustering algorithms have been beneficial, the increasing volume and complexity of malware threats have rendered these techniques less accurate and effective in malware detection. This paper introduces an innovative approach for detecting harmful software within executable files. Our approach diverges significantly from traditional malware detection systems, which typically rely on signatures. Instead, we employ Recurrent Neural Network (RNN) algorithms to differentiate between malicious and benign files. Our experiments demonstrate that our Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) models, followed by traditional clustering algorithms, significantly and effectively aid in discerning between malicious and benign files. This approach benefits from the combination of advanced deep learning techniques and traditional clustering methods. Our experiments show that the GRU model slightly outperformed the LSTM model in terms of accuracy while both GRU and LSTM models perform much better than using only traditional clustering algorithms. The effectiveness of our approach was tested across three different datasets. This work also highlights the potential of deep learning techniques, especially RNN algorithms, in strengthening information security and cyber-security defenses in today's era.