/reinforcement_learning_wf

Attempts to create a website fingerprinting defense using reinforcement learning

Primary LanguagePython

reinforcement_learning_wf

Attempts to create a website fingerprinting defense using reinforcement learning

Note: This project had limited success due to the lack of inherent 'randomness' in DQN and other RL models. To be specific, the WF attacker can use DNNs and access to the RL defense to effectively retrain the attack on defended traffic. Future defenses will have to take this adversarial retraining into account while determining effective strategies.