Efficient Classification of Darknet Access Activity with Partial Traffic

Abstract

AbstractIn this paper we propose a novel approach to classify darknet-access traffic with only partial traffic data, which significantly reduces resource consumption and is as accuracy as prior work. Besides, in order to keep up with the users’ real access activity, we simulate new and old user by simply whether delete the cached consensus document before each access and apply our approach. The experiment results confirm that there does exist a window of cell sequence contributes greatly to distinguish darknet-access traffic. With the window size 75 and the start point 67, we can achieve 95.97% accuracy for new user access scenario. Similarly, with the window size 85 and the start point 44, we achieve 94.43% accuracy for old user access scenario.