A Framework for Measuring Network Neutrality Violations in ISP Networks

Abstract

Network neutrality requires that Internet service providers do not discriminate against any traffic, but the lack of transparency in traffic management ensures that such discrimination is hard to measure and easy to conceal. The paper suggests the use of a multi-faceted approach comprising of active probing, passive monitoring and applications-aware performance measurement as a framework of measuring network neutrality violations in the ISP networks. The framework is meant to detect the discriminative practices, which include throttling, prioritization, blocking, and protocol-specific degradation in heterogeneous access technologies. We design a realistic threat model that also includes the ISP incentives and operation constraints, and derive design requirements that include the focus on accuracy of measurements, scalability, privacy of users, and reproducibility. The suggested architecture unites the concepts of traffic classification, application profiling, and multi-layer QoS and QoE measures to provide the ability to compare fine-grained services, content types, and time intervals. The hybrid measurement methodology is presented and offers a combination of both controlled experiments and large-scale passive flow analysis to enhance coverage and minimize bias. Mechanisms of statistical bias reduction and ground-truth validation are used to separate the areas of deliberate discrimination and performance differences caused by congestion. Empirical evidence has shown that the framework can always identify the presence of unfairness and the existence of performance mismatches in high confidence even when dealing with dynamic networks. The discussion identifies practical deployment, regulatory applicability, and applicability of the encrypted traffic and new transport protocols.