Relationship Privacy Leakage in Network Traffics

With the increasing popularity of social applications, people spend plenty of time in online social networking, such as chatting, sharing daily activities, etc. All these online activities are recorded in social applications and leave fingerprinting in network traffics. Both of these two data sources have the risk of leaking privacy. However, previous studies focus on the privacy leakage in social applications while privacy leakage in network traffics attracts less attention. Moreover, the works on privacy leakage in traffics focus on payload information of the traffics. As the internet is becoming more encrypted, in order to propose a general framework to reveal the privacy problem in the traffics, we turn attention to the traffics without payload information. In this paper, we develop a novel methodology to reveal the risk of relationship privacy leakage in network traffics without payload information. We first extract the chatting flows from the network traffics based on $k$-Nearest-Neighbors method. Then we create the ties between users by matching the chatting flows, and chatting patterns are used to verify whether the matched users are chatting with each other. Finally, network traces are used to evaluate the performance of our methodology. The experiments show we can recognize $96.42\%$ of the chatting flows with the precision of $96.96\%$, and we can detect all the ties between users who are chatting with each other.