TGP-based dynamic traffic camouflage method

In order to address the issue of attackers leveraging traffic analysis to steal users' network behavioral patterns, this paper proposes a dynamic traffic camouflage method based on Twin Gaussian Processes (TGP), referred to as TGP-based dynamic traffic camouflage method (TTC). TTC utilizes Twin Gaussian Processes to capture the pattern changes in the target traffic for feature prediction of the undisguised traffic. After obtaining the predicted features, constraints are applied and traffic reconstruction is performed based on these constrained features. The reconstructed traffic is indistinguish-able from the target traffic, leading to misclassification by attack-ers. Experimental results using the ISCXTor2016 public dataset demonstrate that the proposed TTC method has stronger obfuscation capabilities in traffic classification compared to previous traffic obfuscation methods. It effectively protects user privacy.