UNM researchers publish study on intruder network attacks in Nature Communications

If you wanted to take down a power grid or a network of drones, which part of the system should you attack? Researchers at The University of New Mexico School of Engineering posed the question in hopes of developing better network security mechanisms.

Francesco Sorrentino, professor of Mechanical Engineering, and Amirhossein Nazerian, a now-graduated UNM Ph.D. alumnus, published a study titled "Extreme vulnerability to intruder attacks destabilizes network dynamics," in the journal Nature Communications on the security of cyber-physical systems with a surprising result. Other researchers on the team included Sahand Tangerami of K. N. Toosi University of Technology, Malbor Asllani of Florida State University, David Phillips of UNM, Hernán Makse of City College of New York.

Cyber-physical systems are networks in which cyber components, like software, interact with physical components (hardware, sensors, and actuators) to monitor and control physical processes. For example, networks of autonomous vehicles combine onboard software and hardware with vehicle-to-vehicle interactions to coordinate safe travel and avoid collisions. The research focused on "intruder attacks," in which an intruder has an intact physical component but a compromised cyber component. The cyber components of these networked systems are interconnected, with some agents having many more connections than others. While previous studies have focused on attacks targeting the most highly connected agents, or network hubs, Sorrentino's group found that agents with the fewest connections can make the network more vulnerable overall.

Looking at the problem from the perspective of an attacker, the team wanted to learn how quickly it could destroy the dynamics of a network if one of its agents were compromised. Through mathematical frameworks, they found that agents with fewer incoming connections were more vulnerable than hubs because they have fewer lines of good information coming in. With fewer incoming signals to negate bad information, the agent can be compromised more easily.

The group used power grids as an example of a vulnerable network, among others. These systems employ load shedding to disconnect selected loads when disruptions threaten grid stability, helping to prevent major blackouts. By studying how an attacker could exploit these network-level vulnerabilities, the team aims to develop more effective mechanisms for protecting the grid.

The latest paper relates to work the group published earlier this year suggesting that naturally occurring networks are better at transmitting signals than engineered ones. Future work will examine how to develop networks that can be kept safe in bad environments.

Sorrentino is part of the UNM School of Engineering's National Security Engineering Center, one of five research centers to receive pilot funding this year as the School of Engineering seeks to strengthen cross-disciplinary research in strategic areas.