Skip to Main Content
Distributed detection in wireless sensor networks in the presence of one or more classes of misbehaving nodes is considered. Misbehavior may arise from Byzantine attacks, or may be caused by other factors such as node failure due to hardware or software degradation. We consider binary hypotheses testing where the honest nodes transmit their binary decisions to the fusion center (FC), while the misbehaving nodes transmit fictitious decisions. The FC must identify the misbehaving nodes in order to remove their deleterious effect. We show that each class of nodes can be identified with its operating point on the ROC (receiver operating characteristic) curve. Maximum likelihood estimation of the nodes' operating points is then formulated and solved as an expectation maximization (EM) problem with the nodes' identities as latent variables. The solution from the EM algorithm is then used to classify the nodes and to solve the distributed hypothesis testing problem. Numerical results compared with those from the reputation-based schemes show a significant improvement in both classification of the nodes and hypothesis testing results. We also discuss an inherent ambiguity in the node classification problem and explain how it can be resolved.