Deception Robust Control of Self-Organizing Network of Unattended Sensors

With success of deception control theory, founded recently, in an urban warfare application, we propose a strong case for studying deception in ad-hoc micro-sensor network applications for the use cases with a sentient adversary. We use a Lockheed Martin advanced IR&D work, SONUS (Self-Organizing Networks of Unattended Sensors), as an exemplar that provides robust and scalable solution when random non-adversarial noise is present in the sensor measurements. Since SONUS currently does not account for adversarial noise in its sensor modeling, we present scenarios where a sentient adversary can shape the sensor noise to fool the system into mis-estimating the ground truth and responding with sub-optimal control. We conclude that any distributed system built to operate in the adversarial environment must carefully trade off the need for deception control.