Byzantine Distributed Quickest Change Detection Based on Bounded-Distance-Decoding

Byzantine distributed quickest change detection (BDQCD) is a crucial problem in cyber-physical security. The challenge of this problem is that an AI plus IoT (AIoT) network needs to detect the change as quickly as possible, subject to a false alarm rate, to prevent device damage. In the BDQCD problem, dealing with compromised meters also becomes a challenge since these meters can collaboratively form an attack to lengthen the detection delay in an IoT network. Here we consider the network where a fusion center monitors the occurrence of an abrupt event through a bunch of distributed meters that may be compromised. To solve the challenge, a new coded framework for BDQCD utilizing bounded distance decoding at the fusion center is proposed. First, under sufficient meter-to-fusion-center link capacity, we achieve a theoretical result that the detection delay of our scheme can be asymptotically optimal in special cases. Next, new codebooks under insufficient link capacities are designed based on the theoretical result. Through the simulation results, our coded BDQCDs outperform the state-of-the-art works by achieving significantly shorter detection delay under various attacks.