Preventing traffic analysis for real-time communication networks

We address issues related to preventing traffic analysis in computer networks used for real time mission-critical applications. We consider an IP-based network where headers of packets, including source host address and destination host address, are readable by an observer (i.e., by a potential enemy). Although the encryption of network packets significantly increases privacy, the density of the traffic can still provide useful information to the observer. We take an approach by manipulating traffic in the network through host-based rerouting and traffic padding so that the traffic shows a time-invariant pattern. Thus, the observer can not derive any useful information about the real traffic pattern. By evaluating the performance of the algorithms used for this problem in terms of acceptance rate and execution time, we found that some well-known theoretical optimal and near-optimal algorithms failed to meet one or the other criteria. We present a heuristic method that can effectively prevent traffic analysis while at the same time meeting real-time requirements. Our algorithm generates a plan that specifies where and when the dummy packets should be transmitted and if and how the payload packets should be rerouted and can yield high acceptance rate with low execution time. The success of the algorithm stems from the fact that it explicitly takes into account the real-time requirements and properly balances the traffic over the links.