Robust non-negative matrix factorization for joint spectrum sensing and primary user localization in cognitive radio networks

In this paper, a novel approach based on non-negative matrix factorization is applied for joint spectrum sensing and primary user localization in cognitive radio networks. This approach is robust and tolerant to sparse, yet strong interference caused by malicious attack or false data injection. Simulation results clearly indicate that the proposed method is highly effective in yielding low localization error for various strengths and degrees of sparsity of interferer. It is also shown that the localization performance significantly increases with the number of cognitive radios deployed.