QUIC Traffic Classification Based on Multi-Feature Fusion

Traffic classification is of great significance to network management and security. However, with the QUIC (Quick UDP Internet Connection) full encryption protocol launched by Google, the lack of plaintext information has brought enormous challenges to traffic classification. Existing research primarily focuses on utilizing machine learning methods to explore flow statistical features for the classification of QUIC traffic. However, due to the influence of encryption algorithms, packet size distortion will occur, making the classification accuracy poor and unable to be refined. Given the above problems, this paper proposes a multi-feature fusion classification method. By decrypting the initial packet of QUIC traffic, obtaining the plaintext information of the ClientHello field to extract category features, and integrating numerical flow statistical features and machine learning methods to make QUIC traffic classification more refined and precise. Through the captured QUIC traffic experiment, the fine-grained classification accuracy for QUIC traffic proposed in the paper reaches 99.95%.