DeepCorr - Strong Flow Correlation Attacks on Tor Using Deep Learning

Explore a groundbreaking deep learning-based flow correlation attack on Tor called DeepCorr in this 27-minute conference talk. Delve into the architecture, data collection methods, and comparison with previous works. Examine the impact of time, flow length, and training set size on the attack's effectiveness. Analyze potential countermeasures and experiments on pluggable transports. Gain insights into the implications for Tor's security and the importance of flow correlation in various attacks on the network.

Intro
Flow Correlation Attack
Forcing Traffic Through Malicious Relays
Bottomline: Flow correlation is key to many Tor attacks
Intuition: Tor is super noisy!
Architecture
Data Collection
Comparison with Previous Works
Impact of time
Effect of Flow Length
Countermeasures
Experiments on Pluggable Transports
Conclusions
Size of Training Set
Negligible Circuit Dependency