DeMarking: A defense for network flow watermarking in real-time

Abstract

The network flow watermarking technique associates the two communicating parties by actively modifying certain characteristics of the flow generated by the sender so that it covertly carries some special marking information. Some third-party attackers communicating with the hidden server as a Tor client may attempt de-anonymization attacks to uncover the real identity of the hidden server by using this technique. This compromises the privacy of the anonymized communication system. Therefore, we propose a watermark defense scheme based on deep neural networks. Firstly, we design a training architecture based on generative adversarial networks and adversarial attacks. This architecture can train a converter to convert the original Inter-Packet Delays (IPD) into newly generated “clean” IPDs by the model, causing the adversary’s detector to extract incorrect information and thus unable to perform traffic correlation. Using the trained converter model, we design a watermark defense scheme that can effectively resist time-based watermarking techniques.