Research On QUIC Encrypted Traffic Classification Method Based On Machine Learning

For network service providers and network security regulators,classification of encrypted QUIC traffic can help them provide network services and maintain network security.Due to the characteristics of QUIC protocol,such as full encryption,fast connection establishment,multiplexing,and so on,the existing classification technology of encryption traffic cannot be applied to the classification of encrypted QUIC traffic or the extracted QUIC traffic feature dimension is single,resulting in low classification accuracy.To solve the above problems,this paper proposes a classification algorithm of encrypted QUIC traffic,extracts the characteristics of three dimensions of encrypted QUIC traffic.And it combines with ensemble learning stacking algorithm,constructs a classification model composed of convolutional neural network and naive Bayesian algorithm,to classifies QUIC traffic.Based on the above classification algorithm of encrypted QUIC traffic,this paper designs and implements a classification system of encrypted QUIC traffic,including data acquisition,classification,and data display.The experimental results show that compared with the existing classification algorithm of encrypted QUIC traffic based on a semi-supervised convolutional neural network model,the accuracy of this algorithm is improved by 9%;compared with the previous classification algorithm of encrypted QUIC traffic based on a single convolutional neural network,the accuracy of this algorithm is improved by 7%.