Blind matching algorithm based proxy distribution against internet censorship

Abstract

AbstractProxy based censorship circumvention technique is treated as one of the most promising approaches to fighting against Internet censorship, and its practical applications like the onion router and Lantern have been widely utilized. However, the proxy distribution, as a basic and core issue, still remains unsolved due to (i) the finiteness and non‐renewable nature of proxy resources, (ii) the heterogeneity of clients and proxies requirements, and (iii) the existence of secondary censorship. In this paper, a win‐win driven proxy distribution mechanism for proxy based censorship circumvention systems is designed by leveraging the blind matching theory. Specifically, a combinational manner is developed to measure the quality of service (QoS) requirements of clients in terms of normalized bandwidth and reliability, and security requirements of proxies in terms of credits and suspicion. Then, by introducing the concept of virtual currency, the proxy distribution issue is formulated as two‐sided combinatorial problems from the perspective of clients and proxies, respectively. To solve the complex NP‐hard combinatorial problems in a win‐win manner without unilateral dominance, a two‐sided many‐to‐many context‐free matching game is constructed, in which the aspiration levels and agreement functions are defined to capture the security and QoS concerns of clients and proxies. Further, a modified notion of pairwise stability is exploited as the solution concept, and a client‐proxy blind matching algorithm with market and information decentralization is proposed to obtain the ε‐pairwise stable solution, where clients and proxies make encounters randomly and update their aspiration levels through a stochastic learning process. Finally, theoretical analysis and numerical results are carried out to validate the properties and performance of the proposed algorithm including stability, complexity, and convergence.