Poster: Towards Model Drift Resistant Website Fingerprinting with Time-Series LLMs
Published in 33rd Annual Network and Distributed System Security Symposium (NDSS), 2026
This work investigates model drift in Tor website fingerprinting (WF), including both data drift and concept drift, using a large-scale dataset collected over three months with a cache-enabled Tor Browser. It further explores Time-series Language Models (TSLMs) with CNN-based traffic feature tokenization to improve the robustness and efficiency of WF under model drift.