Adaptive Trust Consensus for Blockchain IoT: Comparing RL, DRL, and MARL Against Naive, Collusive, Adaptive, Byzantine, and Sleeper Attacks

TL;DR

This work tackles securing blockchain-enabled IoT against a spectrum of adversaries by integrating ABAC with Fully Homomorphic Encryption to preserve policy privacy, and by employing learning-based defenses that adaptively manage delegate trust. It conducts a thorough, apples-to-apples comparison of three reinforcement learning paradigms—tabular RL, DRL, and MARL—across five attack families (NMA, CRA, AAA, BFI, TDP) in a 16-node IoT network. The study finds that MARL offers a clear advantage against coordinated attacks like CRA, while DRL generally provides strong, robust detection across most threats; however, Time-Delayed Poisoning remains a critical vulnerability unmitigated by the evaluated approaches. The results highlight the value of coordinated, multi-agent learning for trust defense in decentralized settings and point to the need for longer-term trust memory and temporal anomaly detection to resist patient, trust-building adversaries. The privacy-preserving ABAC-FHE integration further strengthens deployment viability by preventing leakage of attribute and policy information during policy evaluation, enabling secure, adaptive defense in resource-constrained IoT environments.

Abstract

Securing blockchain-enabled IoT networks against sophisticated adversarial attacks remains a critical challenge. This paper presents a trust-based delegated consensus framework integrating Fully Homomorphic Encryption (FHE) with Attribute-Based Access Control (ABAC) for privacy-preserving policy evaluation, combined with learning-based defense mechanisms. We systematically compare three reinforcement learning approaches -- tabular Q-learning (RL), Deep RL with Dueling Double DQN (DRL), and Multi-Agent RL (MARL) -- against five distinct attack families: Naive Malicious Attack (NMA), Collusive Rumor Attack (CRA), Adaptive Adversarial Attack (AAA), Byzantine Fault Injection (BFI), and Time-Delayed Poisoning (TDP). Experimental results on a 16-node simulated IoT network reveal significant performance variations: MARL achieves superior detection under collusive attacks (F1=0.85 vs. DRL's 0.68 and RL's 0.50), while DRL and MARL both attain perfect detection (F1=1.00) against adaptive attacks where RL fails (F1=0.50). All agents successfully defend against Byzantine attacks (F1=1.00). Most critically, the Time-Delayed Poisoning attack proves catastrophic for all agents, with F1 scores dropping to 0.11-0.16 after sleeper activation, demonstrating the severe threat posed by trust-building adversaries. Our findings indicate that coordinated multi-agent learning provides measurable advantages for defending against sophisticated trust manipulation attacks in blockchain IoT environments.

Figures (19)

• Figure 1: Blockchain IoT System Architecture integrating FHE-secured ABAC with trust-based delegated consensus and reinforcement learning defense.
• Figure 2: Five Attack Families: NMA operates independently, CRA coordinates trust manipulation, AAA adapts to defenses, BFI splits consensus, and TDP exploits temporal trust dynamics.
• Figure 3: Reinforcement Learning Interaction: The agent observes network state, selects delegation ratio adjustment, and receives reward based on detection performance and operational metrics.
• Figure 4: Trust-Based Consensus Blockchain: Nodes accumulate trust through valid behavior, with top-ranked nodes forming the consensus committee. Trust scores are updated based on validation outcomes and protocol adherence.
• Figure 5: ABAC integrated with Fully Homomorphic Encryption: Access requests are evaluated entirely on encrypted data, preventing attribute leakage even during policy evaluation.