Behavioral anomaly detection in dynamic self-organizing systems: a cognitive fusion approach

The detection of, and response to, anomalies is one of the prerequisites of implementing system recovery and adaptation mechanisms in dynamic self-organizing systems. In this paper we propose a cognitive fusion approach to the detection of anomalies appearing in the behavior of dynamic self-organizing systems. Our domain of interest includes sensor networks, mobile ad hoc networks, and tactical battle management. The paper argues for three tiers of processes required for behavioral anomaly detection: an event awareness level, a situation awareness level, and a decision awareness level. We discuss self organization in nature and artificial systems and propose an architecture for artificial self-organizing systems that incorporates the three awareness levels. Finally, we propose an application framework in which the architecture could be embodied.