Many Critical Infrastructures (CI) use the Internet as a means of providing services to citizens and for dispatching their own transactions. CIs, like many other organizations connected to the Internet, are prone to cyber-attacks. The attacks can originate from their trusted customers or peer CIs. Distributed network intrusion detection systems (NIDS) can be deployed within the network of national Network Service Providers to support cyber-attack mitigation. However, determining the optimal placement of NIDS devices is a complex problem that should take into account budget constraints, network topology, communication patterns, and more. In this paper we model interconnected CIs as a communication overlay network and propose using Group Betweenness Centrality as a guiding heuristic in optimizing placement of NIDS with respect to the overlay network. We analyze the effectiveness of the proposed placement strategy by employing standard epidemiological models and compare it to placement strategies suggested in the literature.