Peirce’s criteria for Multiple Outlier Detection in Cooperative Spectrum Sensing for Cognitive Radio Network

Abstract

With the advent of number of wireless application the problem of spectrum scarcity is also increasing. Cognitive radio has emerged as the solution to the problem since it exploits the transmission opportunities in the under-utilized spectrum bands of licensed users. A single unlicensed user cannot reliably detect the presence of a licensed user due to numerous channel impairments. Hence cooperative spectrum sensing uses multiple unlicensed users to perform the same task in cognitive radio network (CRN). But CRN suffers from various security risks. One such risk is spectrum sensing data falsification attack. In this attack, an unlicensed user who is malicious will purposely send the false report to fusion centre in the network, there by wrongly influencing the overall decision. The detection and elimination of such malicious user in a CRNcan be done by statistical outlier tests. In this paper, we considerPeirce’s criteria, Dixon’s test, Grubb’s test and Box-plot test as statistical outlier tests. The performance of these tests is compared using the receiver operating characteristics curve. From the simulation results, it is observed that Peirce’s criteria perform better when compared to the rest of the tests for both single as well as the multiple malicious users for all types of spectrum sensing data falsification attacks considered in CRN.