Reproduce “DEMUX: Boundary-Aware Multi-Scale Traffic Demixing for Multi-Tab Website Fingerprinting” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yali Yuan, Yaosheng Liu, Qianqi Niu, Guang Cheng
Researchers developed DEMUX, a framework that identifies multiple concurrently visited websites from interleaved encrypted traffic by addressing three structural requirements for implicit traffic demixing. The system demonstrated improved performance, surpassing previous state-of-the-art methods by up to 9.2 percentage points in specific multi-tab scenarios and showing enhanced robustness against privacy defenses.
Key insights
- Multi-tab website fingerprinting is better understood as an 'implicit traffic demixing problem' rather than a mere extension of single-tab classification.
- Maintaining signal integrity at segment boundaries and modeling traffic patterns across multiple temporal scales are crucial for disentangling interleaved traffic.
- Relative positional encodings (like RoPE) in Transformers are more effective than absolute ones for associating dispersed traffic fragments belonging to the same website in dynamic multi-tab scenarios.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.