Diffusion RLS Algorithm for Secure Distributed Estimation Under Attacks

In this paper, a secure diffusion recursive least-square (S-DRLS) algorithm is proposed to protect the distributed network from attack. Different from the existing diffusion schemes, it consists two steps: (i) weight update (ii) detection and (iii) weight combination, where the second step utilizes non-cooperative RLS (nc-RLS) adaptive filter algorithm to provide reliable reference estimations under attacks and calculate the threshold to detect the trustworthy neighbors of each node. Furthermore, the time-varying threshold selection method is given. Finally, some experiments are performed to show the effectiveness of the proposed algorithm under attacks.