An encrypted traffic classification neural network optimized by heuristic algorithm

As countries around the world have higher and higher requirements for information security, and the traffic transmitted in the network is required to be encrypted to different degrees, so the proportion of encrypted traffic in the network is becoming higher and higher, attack traffic in the network is also becoming encrypted, making it difficult to obtain the type of attack traffic and making defense difficult. However, due to regulatory requirements, network managers want to understand the distribution of different types of traffic or attack traffic in the network without decryption, so the classification of encrypted traffic has become a difficult problem. Aiming at tackling this challenge, this study provides a method based on heuristic algorithm to do feature optimization, paired with deep learning for classification, and the accuracy of classification is used as the optimization equation of heuristic algorithm. This experiment leverages the ISCX-VPN2016 public dataset for experimental study. the final experimental accuracy has reach to 94.17%, the model displays good optimization effect, and the model can be transplanted to other tasks, with great robustness.