Deep Nearest Neighbor Website Fingerprinting Attack Technology

By website fingerprinting (WF) technologies, local listeners are enabled to track the specific visited by users through an investigation of encrypted traffic between and Tor network entry node. The current triplet (TF) technique proved possibility small sample WF attacks. Previous research methods only concentrate on extracting overall features while ignoring importance characteristics for Thus, in present paper, a deep nearest neighbor (DNNF) attack technology is proposed. websites extracted via convolutional neural (CNN), then k-nearest (k-NN) classifier utilized classify prediction. When provides 20 samples, accuracy can reach 96.2%. We also found that DNNF method acts well compared traditional coping with transfer learning concept drift problems. In comparison TF method, classification proposed improved 2%–5% it dropped 3% when classifying data collected from same after two months. These experiments revealed more flexible, efficient, robust technology, particularly important