A method and system for preventing encrypted traffic fingerprinting

Abstract

This application relates to a method and system for encrypted traffic fingerprinting defense, comprising: real-time capture of network traffic of a target app; extraction of burst-level traffic features from the network traffic of the target app, representing them as an app feature matrix; measurement of traffic similarity of the app feature matrix; allocation of the app feature matrix to different anonymity sets, and merging of the anonymity sets to obtain an scalable anonymity set; construction of a dedicated conditional generative adversarial network model, and collection of traffic samples to train the conditional generative adversarial network model, thereby generating obfuscated traffic trajectories; optimization of the obfuscated traffic trajectories using a dynamic optimization strategy to form obfuscated traffic adapted to the actual network environment, and restoration of the obfuscated traffic to the original traffic before forwarding it to the destination server. Using the technical solution of this application, similar apps can be accurately aggregated, matching obfuscated traffic can be generated, network adaptation and edge restoration can be achieved, balancing encrypted traffic defense, network performance and privacy security.

Images