Consensus-Based Distributed Quickest Detection of Attacks With Unknown Parameters

Sequential attack detection in a distributed sensor network is considered, where each successively produces one-bit quantized samples of desired deterministic scalar parameter corrupted by additive noise. The unknown parameters the pre-attack and post-attack models, namely to be estimated injected malicious data at attacked sensors pose significant challenge for designing computationally efficient scheme detect occurrence attacks only using local communication with neighboring sensors. generalized Cumulative Sum (GCUSUM) algorithm which replaces their maximum likelihood estimates CUSUM test statistic. For problem under consideration, sufficient condition provided expected false alarm period GCUSUM can guaranteed larger than any given value. Next, we consider implementation GCUSUM. We first propose an alternative statistic asymptotically equivalent that Then based on proposed running consensus algorithms, approximate significantly reduce prohibitively high computational complexity centralized Numerical results show provide performance comparable