Distributed Detection of Information Flows

Distributed detection of information flows is considered in which traffic sensors at different locations of a network observe transmission epochs. The traffic sensors communicate their measurements to a fusion center via channels with rate constraints, and the fusion center performs hypothesis testing for information flow detection. Under a nonparametric flow model where relayed packets can be perturbed up to bounded delays and multiplexed with chaff noise, flow detectability is characterized through a notion called consistency-rate function that shows the level of detectable flows under capacity constraints on the fusion channels. Achievability results are presented by constructing detection systems consisting of quantization, data transmission, and detection subsystems. In particular, slot-by-slot quantization schemes at the local sensors and threshold detection schemes at the fusion center are proposed to provide consistent detection with quantifiable performance.