Network Obfuscation for Net Worth Security

Abstract

Research community has been highlighting recently security issues related to network traffic analysis. Using machine learning techniques, an eavesdropper can exploit traffic features to determine useful information threatening privacy of users. In this paper, we propose an efficient defense against network fingerprinting attacks where we obfuscate information leaked by traffic features specifically packet sizes. First, we model the packet lengths probability distribution of the source app to be protected and that of the target app that the source app will resemble. Then, we define a security model that mutates the packet lengths of a source app to those lengths from the target app having similar bin probability. This would confuse a classifier and make it identify a mutated source app as the target app. A comprehensive simulation study of the proposed model, using real apps traffic traces, shows considerable obfuscation efficiency with relatively acceptable overhead. We were able to reduce a classification accuracy of 91.1% to 0.22% using the proposed algorithm, with only 11.86% padding overhead.