Secrecy Capacity Scaling of Large-Scale Networks With Social Relationships.

Due to the nature of a wireless medium, a wireless channel is susceptible to malicious nodes, which encourages the consideration of secrecy in many applications such as satellite networks, social networking services, etc. Motivated by the social characteristic that makes nodes more likely to communicate with nearby neighbors, we study the secrecy capacity of large-scale wireless networks with social relationships. We consider two distributions of legitimate nodes: They are homogeneously placed based on a Poisson point process (PPP), or they are inhomogeneously distributed as multiclustering topology. In particular, we consider two scenarios: 1) the noncolluding case, where each eavesdropper decodes its receiving message independently; and 2) the colluding case, where eavesdroppers can cooperate to decode the message. We adopt a rank-based model to describe the social characteristic and investigate its impact on the secrecy capacity using the technique of self-interference cancelation. We employ a passive attack model, where eavesdroppers can only listen to transmissions among legitimate nodes. Our important findings include the following: 1) In the noncolluding case, secrecy capacity is not influenced by eavesdroppers in order sense, but it varies with the density of eavesdroppers and social impacts in the colluding case. 2) The proposed scheme can achieve optimal secrecy throughput neglecting polylogarithmic factor in homogeneous wireless networks. [ABSTRACT FROM PUBLISHER]