Defending against Deep-Learning-Based Flow Correlation Attacks with Adversarial Examples

Table 1

The related work of flow correlation attack, flow correlation defense, and website fingerprint defense.

Scheme	Method	Innovation points	Authors	Drawbacks
Flow correlation attack	Watermark attacks	Modified the packet flows to “fingerprint” them.	Shmatikov et al. [5]	Require high privileges and break the original communication easily.
	Timing based	Use the traffic patterns to correlate flows.	Paxson and Zhang [6]	Low accuracy.
	Bayesian traffic analysis	Developed Bayesian traffic analysis techniques to process sampled data.	Murdoch and Zieliński [8]	Cannot correlate lots of short-lived connections.
	Fine-grained level detection	Correlated the aggregate sizes of network packets over time.	Blum et al. [7]	Low accuracy.
	Asymmetric traffic analysis	Further combined the asymmetric traffic analysis and BGP hijacking to deanonymize users	Sun et al. [9]	Only useful for BGP hijacking.
	Deep learning based	Use CNNs models to learn a flow correlation function and achieve drastically higher accuracies.	Nasr et al. [10]	Require hardware support.
Flow correlation defense	Counter-RAPTOR	Reduced the chance of adversary observed network traffic.	Sun et al. [11]	Only useful for defending BGP hijacking.
	Obfs4	Randomly obfuscate packets time and size.	Tor project. [12]	Unacceptable bandwidth overhead.
	ScrambleSuit	Use morphing techniques.	Winter et al. [43]	Unacceptable bandwidth overhead.
Website fingerprint defense	Application layer defense	Changed the routing algorithm or confused HTTP requests.	Wladimir et al. [13] Giovanni et al. [14] Henri et al. [15]	Hard to implement in real world.
	Network layer defense	Fool the classification model by inserting dummy packets.	Juarez et al. [19] Wang et al. [20]	Cannot defend the deep-learning-based attack.