Distributed Detection of Multi-Hop Information Flows With Fusion Capacity Constraints

Abstract

The problem of detecting multihop information flows subject to communication constraints is considered. In a distributed detection scheme, eavesdroppers are deployed near nodes in a network, each able to measure the transmission timestamps of a single node. The eavesdroppers must then compress the information and transmit it to a fusion center, which then decides whether a sequence of monitored nodes are transmitting an information flow. A performance measure is defined based on the maximum fraction of chaff packets under which flows are still detectable. The performance of a detector becomes a function of the communication constraints and the number of nodes in the sequence. Achievability results are obtained by designing a practical distributed detection scheme, including a new flow finding algorithm that has vanishing error probabilities for a limited fraction of chaff packets. Converse results are obtained by characterizing the fraction of chaff packets sufficient for an information flow to mimic the distributions of independent traffic under the proposed compression scheme.