Resilienta nätverkade system i fientliga miljöer

Game theory provides a set of powerful methodologies and concepts for the design of scalable algorithms for resource management in multi-agent networked system.

Yet, existing game theoretical methodologies and concepts do not address the potential impact of targeted attacks against the constituent nodes and the communication between those, with potentially devastating attacks as a consequence.

In this project we aim at developing an understanding of the vulnerabilities of multi-agent networked systems and their impact, through addressing fundamental problems related to finding equilibria and equilibrium performance under adversarial behavior.

We will develop models of the interaction between adversarial and benign autonomous nodes, and will develop computationally tractable game theoretical models for analyzing various learning algorithms and the emergence of equilibria in the presence of adversarial behavior.

Finally, we will use the developed models for the design of topologies and learning algorithms that are resilient to adversarial behavior and allow to reach good equilibrium performance.

The project addresses an area of game theory that has received little attention so far, and the developed results could find application in critical networked systems, from autonomous vehicular systems through federated mobile edge clouds to social networks and future smart power grids.