A Neural Network Based User Identification for Tor Networks: Comparison Analysis of Different Activation Functions Using Friedman Test
Tetsuya Oda, Ryoichiro Obukata, Masafumi Yamada, Masahiro Hiyama, Leonard Barolli, Makoto Takizawa · 2016
In this paper, we present the application of Neural Networks (NNs) for user identification in Tor networks. We used the Back-propagation NN and constructed a Tor server, a Deep Web browser (Tor client) and a Surface Web browser. Then, the client sends the data browsing to the Tor server using the Tor network. We used Wireshark Network Analyzer to get the data and then used the Back-propagation NN to make the approximation. For evaluation we considered Number of Packets (NoP) metric and activation function. We analyze the data using Friedman test. From the results, we adopt null hypothesis H0since p <; 0.05 for all activation functions. However, the softsign/x has the smallest p-value among activation functions. Therefore, it is better to use softsign/x for bad user identification in Tor networks.