Adversarially Robust Blockchain Reputation System for Additive Manufacturing Supply Chains

Additive manufacturing (AM) supply chains comprise independent organizations whose reliability directly affects product quality. Existing blockchain-based provenance systems provide immutable traceability, but neither measure participant trustworthiness nor incentivize responsible behavior. This paper presents a blockchain-integrated reputation framework for AM supply chains, implemented on Hyperledger Fabric, that combines stake-based economic security with a Bayesian reputation model. The system incorporates five complementary security controls stake requirements, self-rating prevention, role-based access control, duplicate-rating protection, and cryptographic evidence hashing, and couples them with a multi-dimensional Bayesian scoring method based on Beta distributions. The design supports constant-time updates, Wilson confidence bounds for uncertainty quantification, and temporal decay to emphasize recent behavior. We evaluate the system through performance benchmarking and adversarial testing. The implementation achieves sequential read and write latencies of 6.63 ms and 99.86 ms, concurrent read throughput of 612.94 TPS (Transaction Per Second), and concurrent write throughput of 113.48 TPS. Under contention, the network correctly enforces MVCC (Multi-Version Concurrency Control) isolation by rejecting 99.0% of conflicting transactions while sustaining 1.54 TPS. Storage overhead remains low at approximately 220 bytes per transaction. Security validation shows an 83.3% prevention rate across mechanism-level adversarial tests, and economic attack simulations indicate improved Sybil resistance, with the Bayesian model exhibiting a 14% reputation shift versus 30% for an averaging baseline. Overall, the results demonstrate a practical and adversarially validated approach to decentralized trust in AM supply networks.