- The paper presents an interrogator module that integrates continuous behavioral trust assessment into decentralized IoUT agent coordination, achieving 94.2% anomaly detection accuracy.
- It employs a quantized transformer encoder and permissioned blockchain governance to secure multi-agent communications under tight energy and bandwidth constraints.
- Experimental evaluations demonstrated enhanced communication reliability, scalability, and minimal energy overhead compared to static and Bayesian trust models.
Interrogator-Based Secure Framework for Decentralized IoUT Agents
Introduction and Motivation
The proliferation of Autonomous Underwater Vehicles (AUVs) and sensor nodes within Internet of Underwater Things (IoUT) deployments necessitates robust frameworks for secure, reliable, and autonomous multi-agent coordination under severe environmental and operational constraints. Most prior architectures rely on static trust assumptions post-authentication, which exposes deployments to persistent vulnerabilities due to behavioral compromise or credential misuse. This paper introduces a framework that integrates behavioral trust assessment directly within decentralized IoUT coordination, emphasizing continuous, passive, and scalable monitoring compatible with underwater acoustic communication constraints.
Architectural Design and System Modeling
The proposed architecture is structured into four interacting layers: the operational underwater agents, a behavioral monitoring layer, a permissioned blockchain-based governance tier, and surface infrastructure for validation and oversight. The interrogator module, a privileged software layer, passively analyzes communication metadata (e.g., inter-packet timings, retransmissions, routing stability) using a quantized transformer encoder with minimal computational and energy overhead. This module computes dynamic trust scores and manages conditional access to mission-critical data flows.
Operational autonomy is preserved by decoupling trust evaluation from sensory decision processes. Agents operate within an observe-decide-act cycle, optimizing cumulative expected reward over mission horizons, factoring sensing fidelity, routing reliability, and energy sustainability under strict bandwidth and power constraints.
Behavioral Trust Inference and Enforcement
The interrogator utilizes transformer-based sequence modeling to infer trustworthiness based on behavioral metadata, prioritizing energy efficiency by minimizing payload analysis. Trust values are updated via exponential smoothing, and enforcement actions are triggered when scores drop below preset thresholds. The transformer architecture outperformed quantized LSTM baselines in stability and inference latency, especially under sparse acoustic traffic characteristic of underwater communication.
Permissioned blockchain (Hyperledger Fabric, PBFT consensus) stores agent identity and behavioral records securely and efficiently, mitigating consensus-induced latencies by aggregating trust deltas and only logging confirmed security events. Tiered enforcement mechanisms allow immediate local isolation or transmission throttling in response to anomalies, with escalated actions validated by consortium governance.
Threat Model and Adaptive Security
The architecture defends against classic IoUT adversaries—credential theft, packet injection, replay attacks, insider threats—by leveraging redundant cross-verification and temporal pattern analysis, raising stealth evasion costs. Security adaptation is dynamically modulated: monitoring intensity increases with anomaly detection, while false positives induced by environmental variation are reduced via persistence checks and smoothing.
Incident response is staged: agents exhibiting anomalous behavior are excluded from routing, neighbors recompute relay paths, and persistent deviators are isolated. This containment strategy ensures rapid mitigation without significant disruption to network continuity or operational autonomy.
Experimental Evaluation
Prototype validation was conducted with NS-3 Aqua-Sim, simulating 50 heterogeneous agents within a 1 km² area using 10–20 kbps acoustic links and realistic latency models. The transformer-based interrogator achieved 94.2% mean anomaly detection accuracy, surpassing static (72.5%) and Bayesian reputation (86.1%) baselines, with early-stage detection latency reduced by 27%. Communication reliability was enhanced, with packet delivery ratios reaching 91.6%, compared to 79.4% (static) and 86.7% (Bayesian) models. Energy overhead was a modest 5.8% increase relative to static trust solutions, staying within the operational energy budgets for AUVs. Detection accuracy and enforcement latency remained stable as agent count increased, evidencing the scalability of the approach.
Practical and Theoretical Implications
This framework supports resilient underwater multi-agent deployments by moving beyond static credential-based trust, integrating temporal behavioral monitoring with decentralized governance. Practical implications include improved reliability for environmental monitoring, maritime security, and infrastructure inspection tasks, especially on long-duration missions where autonomy is paramount and human intervention is minimal. The dynamic behavioral scoring and permissioned ledger foundation also enable zero-trust operational models, raising the security baseline for distributed edge AI in IoUT scenarios.
Theoretically, the architecture advances the co-design of behavioral inference and decentralized governance under real-world communication and energy constraints, demonstrating the feasibility of embedding lightweight ML models for trust evaluation directly into constrained cyber-physical systems.
Future Directions
Prospective developments involve physical deployment validation, optimization of transformer architectures for ultra-low-power underwater platforms, and adaptation to intermittent acoustic links and variable consensus configurations. Addressing adversarial adaptation to behavioral monitoring and further scaling consortium ledger architectures to vast agent networks are also critical directions.
Conclusion
The interrogator-based secure framework for IoUT demonstrates significant improvements in anomaly detection, communication reliability, and operational efficiency through continuous behavioral trust validation. By decoupling operational control from trust evaluation, leveraging transformer-driven inference, and enforcing permissioned blockchain-based governance, the architecture enables secure, scalable, and autonomous underwater multi-agent coordination. Continuous behavioral verification supersedes static trust assumptions, establishing a robust foundation for secure IoUT deployments under adversarial and resource-constrained maritime conditions (2604.04262).
