Trust management for underwater Internet of Things: A combined hidden Markov and cloud model approach

Open Webpage

Fangyuan Xing, Zhiquan Jiang, Yutian Tan, Fei Tong

Published: 2025, Last Modified: 08 Nov 2025Ad Hoc Networks 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Underwater wireless optical sensor networks (UWOSN) have emerged as a transformative solution for marine exploration systems, revolutionizing applications ranging from seabed mineral prospecting to real-time oceanic monitoring through their unparalleled bandwidth and ultra-low latency. However, as underwater sensors and devices tend to persistently operate in unattended harsh environments, they are vulnerable to attacks by security threats, including hybrid attacks targeting both network nodes and communication paths. Due to the limitations in node energy and channel characteristics, traditional encryption algorithms and trust management approaches designed for terrestrial wireless sensor networks (WSN) are unsuitable for underwater WSN. To this end, this paper studies trust management for UWOSN considering both network node and link attacks. Aligned with clustered network architectures, a three-layer trust management framework is proposed in this paper. The collection layer harvests node and link trust evidences, the processing layer computes node and link comprehensive trust clouds, and the decision-making layer determines node and link states via hidden Markov model for final trust adjudication. Hidden Markov model and cloud model are combined and employed to eliminate discrepancies between the observed states inferred from the collected trust evidences and the hidden states showing the actual states of UWOSN nodes and links in dynamic environments. Simultaneously, recognizing that uniform threshold standards are unsuitable for trust management of sensor nodes in geographically marginalized positions, the proposed dynamic thresholds are designed to overcome spatial constraints and mitigate elevated false positive rates. Experimental results demonstrate that the proposed combined hidden Markov and cloud model (HMCM) trust management approach outperforms related notable studies in terms of detection rate, false detection rate, and packet delivery rate.