Encrypted Network Traffic Classification and Feature Selection by Ensemble of CNN and TLBO Meta-heuristic Algorithm

The network traffic and its classification are crucial to network administration and monitoring. The extensive use of encryption techniques and the dynamic ports policy make it difficult for standard traffic classification algorithms to classify encrypted data. Deep learning techniques have lately been the subject of in-depth research for network traffic categorization. Unfortunately, a lot of training data is needed for these models. The fact that the characteristics for most traffic categorization algorithms must be retrieved by a specialist presents another difficulty. Finding the required elements that contribute to a better categorization using these approaches is highly laborious and time-consuming. In order to construct a traffic classification model that properly identifies traffic categories, this study combines the convolutional neural network (CNN), teaching–learning-Based optimization algorithm (TLBO), and self-organizing maps (SOM) methods. The suggested approach is a blend of CNN and TLBO, where CNN is an image-based technique that can accurately identify encrypted network data and TLBO is a feature selection mechanism that combines self-organizing maps (SOM). This approach is highly lightweight and has the ability to automatically extract features, choose features, and categorize encrypted network information.