Pars Anonymity Network Traffic Flow Analysis Using Machine Learning

Anonymity is one of the fundamentals of privacy in the internet that should be strictly considered by governments and ISPs. Network traffic flow detection, is considered as detecting the nature of this traffic; Thus, if the traffic of an anonymizer is detected, it means that classified data is being transmitting throw the network, which in return is a great flaw in the anonymity system. Traffic classification - which has various applications - is one of the most powerful methods in datamining. Traffic management via detecting network traffic flow, is viewed as one of these applications. In this research, by using datamining techniques, in the first step the detection rate of Pars Anonymizer (as a domestic anonymizer) is assessed in compare with The Onion Router, Invisible Internet Project, JonDo and HTTPS Traffic, and at the next step, in a more detailed way, the classification rate of four different services in the desired anonymizer was studied. Results suggest that the classification accuracy rate of these experiments at the first step is 100% and at the next step -with the use of Random Forest algorithm- is 95%. In addition, by evaluating the used specifications in every experiment, the effectiveness of these specifications on the overall accuracy and the model build time was assessed.