Vision Transformer and Improved Bi-Directional Long Short-Term Memory based Secure Network Traffic Analysis and Threat Detection

In these recent years, the rapid growth of network traffic has witnessed due to increased complexity of cyberattacks. Besides, detecting network traffic threat and anomalies in real-time is essential for network security enhancement. However, it has high false positive rates in security and resources that leads to unnecessary diverting attention from real threats. Therefore, in this research Vision Transformer (ViT) with Improved Bi-directional Long Short-Term Memory (IBi-LSTM) is proposed for detecting the threats in Real-time. Initially the input data is collected from Network traffic dataset and preprocessing is done through Min-Max Scaling. Then the relevant features based on traffic patterns and anomalies which are extracted through ViT. Finally, for detecting the threats in real time the LSTM with the Adam optimizer is performed. From the results, the proposed IBi-LSTM model achieved better output when compared to existing established better results than existing LSTM model with an accuracy (acc) of 99.85%, specificity of 99.95%, sensitivity 99.99% respectively.