Tor Anonymous Traffic Identification Based on Parallelizing Dilated Convolutional Network

The widespread use of the onion browser (Tor) has provided a breeding ground for proliferation cybercriminal activities and Tor anonymous traffic identification method been used to fingerprint web identify websites visited by illegals. Despite considerable progress in existing methods, problems still exist, such as high training resources required model, bias features due fast iteration singularity definition direction features. On this basis, model based on parallelizing dilated convolutions multi-feature analysis proposed paper order address these perform better website fingerprinting. A single-sample augmentation data combining multi-layer RBMs are performed, binary classification multi-classification conducted different scenarios. Our experiment shows that recognition achieves 94.37% accuracy gains significant drop time both closed-world open-world At same time, enhanced enhance robustness generalization our model. With techniques, efficiency improved we able achieve advantage bi-directional deployability communication link.