Identification and Infiltration in Consensus-type Networks*

This paper examines the system dynamics of a controlled networked multi-agent system, operating with a consensus-type algorithm, that is under the influence of attached node(s). We introduce an identification scheme, involving excitation and observation of the network by the attached node(s), that identifies the spectrum of the underlying system matrix - in this case the influenced network's modified graph Laplacian. Following this identification, infiltration node(s) are then attached to a set of nodes in the network with the objective of sabotaging the network by delivering a constant mean control signal. The spectrum of the modified graph Laplacian provides bounds on the convergence of the system states due to infiltration, quantifying the network's security. We also derive bounds on the effectiveness of the network infiltration in a more general setting by examining the controllability gramian of the infiltrated consensus-type coordination algorithm.