ANGP: A Decentralized Reputation-Based Consensus Protocol for Directed Acyclic Graphs

Abstract

The Adaptive Neural Gossip Protocol (ANGP) is a fully decentralized, asynchronous consensus mechanism designed for Directed Acyclic Graph (DAG) based distributed ledgers. Unlike classical Byzantine Fault Tolerant (BFT) systems that rely on leader election or quorums, ANGP uses:• Amedian-based consensus computed from predictions exchanged via gossip.• A continuous reputation engine that distinguishes honest nodes from Byzantine attackers, including coordinated collusion, rare attacks, sensor faults, and network impairments.• Alightweight Proof-of-Work (PoW) layer (SHA-512/256) to prevent Sybil identity floods, while keeping the core protocol free of staking or token-based governance.ANGP tolerates up to 44% coordinated attackers and 66% uncoordinated attackers with no degradation in honest node safety. It operates asynchronously, requires no global time synchronization, and self-heals under packet loss and network delays. This document provides the complete architectural blueprint, component specifications, security analysis, and integration guidelines for building a production-grade DAG based cryptocurrency or distributed application on top of ANGP.