Distributed Detection of Information Flows

    T. He and L. Tong

    April, 2007.

    Abstract
    This paper addresses the detection of information flows through a
    pair of nodes in a wireless ad hoc network by correlating the
    transmission timing of these nodes. The transmission timing is
    subject to perturbations such as random delays and reshuffling,
    and there may be chaff traffic induced by traffic multiplexing at
    intermediate nodes or the active insertion of dummy traffic.
    Furthermore, the collection of timing measurements is subject to
    capacity constraints. Under the assumption that information flows
    satisfy a bounded delay constraint, the paper characterizes the
    tradeoff between data collection payload and detection quality.
    The detection performance is measured by the maximum amount of
    chaff traffic that allows consistent detection. An upper bound on
    the maximum amount of chaff is derived by formulating the problem
    as a rate-distortion problem. Joint compression and detection
    schemes are proposed for detecting information flows with and
    without full side-information, respectively. The proposed schemes
    are shown to achieve consistent detection in the presence of chaff
    traffic of positive rate under arbitrarily small capacity
    constraints. The proposed detectors are proved to be optimal for
    the corresponding compression schemes, and a comparison of their
    performance gives novel insights into the quantizer design.