Graph modelled system change detection in WSNs

Graphs are suitable to model topology and data patterns in systems such as WSNs. To detect change, there is a need for graph comparison, a computationally demanding task difficult to run on constrained devices. For monitoring, the definition of normal patterns, and deviation from normal are required. In this contribution a flexible graph comparison method allowing monitoring of normal patterns and metrics providing measures of deviation from normal are proposed. In this manuscript, we apply the method to the system modelled by synthetic and random graphs. We demonstrate that the fingerprints of normal topology and data patterns can be acquired with the measures of deviation from normal. We discuss applicability of the method at the edge of WSNs.