GOSSIP-BASED TRUST AND REPUTATION MECHANISM FOR BROKER INTERACTION IN NETWORK-CENTRIC DISTRIBUTED COMPUTER SYSTEMS

Abstract

The object of this research is the broker interaction layer of Network-Centric Distributed Computer Systems (DCS), in which autonomous peer-to-peer brokers replace centralized control planes to manage resource allocation and task scheduling. The problem addressed is the absence of a decentralized trust mechanism capable of detecting and isolating malicious or Byzantine brokers (nodes that silently drop tasks, falsify resource availability, or engage in collusion) without reintroducing the single points of failure that the network-centric architecture was designed to eliminate. To resolve this, a gossip-based trust and reputation mechanism is proposed and integrated directly into the NCA broker interaction layer. The mechanism introduces a composite reputation vector comprising three task-centric metrics (Task Success Rate, Task Delay Factor, and Fault Tolerance Score), coupled with an exponential temporal decay function to rapidly detect oscillating malicious behavior. A credibility-weighted reconciliation rule neutralizes localized collusion by weighting incoming gossip reports against the reporter's own trust score. These trust states serve as a zero-latency pre-selection filter for the task scheduler. The results are explained by the logarithmic convergence of the push-pull epidemic protocol, which propagates trust updates across all honest brokers in O(log N) rounds without centralized coordination. Empirical simulations confirm that the mechanism sustains over 93% task completion rates under 20% Byzantine fault injection and reduces makespan by approximately 25%, generating fewer than two additional gossip messages per scheduled task. The mechanism can be applied in any distributed computing environment operating a horizontal broker control plane, particularly in edge computing, IoT, and cloud federation scenarios where centralized security authorities are impractical.

Keywords: Distributed Computer Systems, Network-Centric Architecture, Gossip Protocol, Trust and Reputation Management, Byzantine Fault Tolerance, Broker Interaction, Epidemic Dissemination