Abstract-This paper addresses the problem of resilient innetwork consensus in the presence of misbehaving nodes. Secure and fault-tolerant consensus algorithms typically assume knowledge of nonlocal information; however, this assumption is not suitable for large-scale dynamic networks. To remedy this, we focus on local strategies that provide resilience to faults and compromised nodes. We design a consensus protocol based on local information that is resilient to worst-case security breaches, assuming the compromised nodes have full knowledge of the network and the intentions of the other nodes. We provide necessary and sufficient conditions for the normal nodes to reach asymptotic consensus despite the influence of the misbehaving nodes under different threat assumptions. We show that traditional metrics such as connectivity are not adequate to characterize the behavior of such algorithms, and develop a novel graph-theoretic property referred to as network robustness. Network robustness formalizes the notion of redundancy of direct information exchange between subsets of nodes in the network, and is a fundamental property for analyzing the behavior of certain distributed algorithms that use only local information.

This paper addresses the problem of resilient consensus in the presence of misbehaving nodes. Although it is typical to assume knowledge of at least some nonlocal information when studying secure and fault-tolerant consensus algorithms, this assumption is not suitable for large-scale dynamic networks. To remedy this, we emphasize the use of local strategies to deal with resilience to security breaches. We study a consensus protocol that uses only local information and we consider worst-case security breaches, where the compromised nodes have full knowledge of the network and the intentions of the other nodes. We provide necessary and sufficient conditions for the normal nodes to reach consensus despite the influence of the malicious nodes under different threat assumptions. These conditions are stated in terms of a novel graphtheoretic property referred to as network robustness.

In the past decade, numerous consensus protocols for networked multi-agent systems have been proposed. Although some forms of robustness of these algorithms have been studied, reaching consensus securely in networked multi-agent systems, in spite of intrusions caused by malicious agents, or adversaries, has been largely underexplored. In this work, we consider a general model for adversaries in Euclidean space and introduce a consensus problem for networked multi-agent systems similar to the Byzantine consensus problem in distributed computing. We present the Adversarially Robust Consensus Protocol (ARC-P), which combines ideas from consensus algorithms that are resilient to Byzantine faults and from linear consensus protocols used for control and coordination of dynamic agents. We show that ARC-P solves the consensus problem in complete networks whenever there are more cooperative agents than adversaries. Finally, we illustrate the resilience of ARC-P to adversaries through simulations and compare ARC-P with a linear consensus protocol for networked multi-agent systems.

Recently, many applications have arisen in distributed control that require consensus protocols. Concurrently, we have seen a proliferation of malicious attacks on large-scale distributed systems. Hence, there is a need for (i) consensus problems that take into consideration the presence of adversaries and specify correct behavior through appropriate conditions on agreement and safety, and (ii) algorithms for distributed control applications that solve such consensus problems resiliently despite breaches in security. This paper addresses these issues by (i) defining the adversarial asymptotic agreement problem, which requires that the uncompromised agents asymptotically align their states while satisfying an invariant condition in the presence of adversaries, and (ii) by designing a low complexity consensus protocol, the Adversarial Robust Consensus Protocol (ARC-P), which combines ideas from distributed computing and cooperative control. Two types of omniscient adversaries are considered: (i) Byzantine agents can convey different state trajectories to different neighbors in the network, and (ii) malicious agents must convey the same information to each neighbor. For each type of adversary, sufficient conditions are provided that ensure ARC-P guarantees the agreement and safety conditions in static and switching network topologies, whenever the number of adversaries in the network is bounded by a constant. The conservativeness of the conditions is examined, and the conditions are compared to results in the literature.

No abstract