A Two-Hop Neighborhood Based Berserk Detection Algorithm for Probabilistic Model of Consensus in Distributed Ledger Systems

Emerging distributed ledger technologies are often not based on Proof of Work (PoW) or Proof of Stake (PoS) consensus protocols. The lightweight protocols, based on the voter model, are typically used for handling contention. However, such protocols are fraught with a particular type of Byzantine adversary known as Berserk adversaries who intend to break the consensus. The existing method of Berserk detection involves the exchange of signatures. This in turn requires key servers, subject to a single point of failure. This paper investigates a new method of Berserk detection. Unlike most of the existing deterministic detection methodologies, the proposed method does not use signatures for the detection of Berserk behavior. The proposed solution is based on two-hop neighborhood opinion information gathering and detects Berserk nodes with some degree of certitude. We also try to ensure that the proposed approach detects most of the Berserk nodes and at the same time keeps the number of false detections marginal.