Towards Lightweight User Identification of Anonymous Cryptocurrency Wallet via Encrypted Traffic Correlation
Xiangdong Kong, Jizhe Jia, Jinhe Wu, Meng Shen, Liehuang Zhu · 2024
With the widespread use of cryptocurrencies and the development of anonymity network technology, how to effectively identify cryptocurrency transactions through anonymity networks such as Tor has become a major challenge in cybersecurity. We introduce a new traffic correlation technique, TSMCorr, aimed at identifying cryptocurrency transactions through anonymous networks like Tor. Traditional traffic correlation methods struggle with the high cost of deployment, while we leverage advanced feature engineering and deep learning, including a Traffic Volume Matrix (TSM), to develop a more accurate and efficient flow correlation model. TSMCorr not only improves upon existing methods in terms of F1 score by $15.5 \%$ on DeepCoFFEA dataset, but also lowers the computational time by $89 \%$, RAM consumption by $77.4 \%$, and model parameters by $11.5 \%$.