Robust collaborative spectrum sensing in the presence of deleterious users

Abstract

Collaborative spectrum sensing has attracted significant research attention in the last few years and is widely accepted as a viable approach to improve spectrum sensing reliability. Fusing data from multiple opportunistic users (OUs) in order to produce reliable sensing results implies a reliance on the OU to provide correct information. In the presence of malfunctioning or selfish users, performance of collaborative spectrum sensing deteriorates significantly. In this study, the authors propose mechanisms for the detection and suppression of such deleterious OUs (DOUs) for hard and soft decision fusion. More specifically, a credibility-based mechanism for hard decision fusion using a hard decision combining beta reputation (HDC-BR) system is introduced. The authors proposed method does not require knowledge of the total number of deleterious users in advance. In HDC-BR, the fusion centre assigns and updates weights to each user's decisions based on an individual user credibility score, which is calculated using the BR system. The presence of DOUs in soft decision-based collaborative spectrum sensing has even more adverse effects on system performance. The authors also propose a scheme for the case of soft decision fusion to detect and eliminate falsified user observations at the fusion centre using a modified Grubbs test; they refer to it as soft-decision combining-modified Grubbs (SDC-MG). They compare the performance of the proposed methods with malicious user detection schemes proposed in the literature as well as with the case where no DOU suppression scheme is implemented, and conclude that SDC-MG performs much better than HDC-BR in a low signal-to-noise ratio regime.