Energy-Efficient Hierarchical Federated Anomaly Detection for the Internet of Underwater Things via Selective Cooperative Aggregation

Kenechi Omeke, Michael Mollel, Lei Zhang, Qammer H. Abbasi, Muhammad Ali Imran · 2026

Read Energy-Efficient Hierarchical Federated Anomaly Detection for the Internet of Underwater Things via Selective Cooperative Aggregation by Kenechi Omeke, Michael Mollel, Lei Zhang, Qammer H. Abbasi, Muhammad Ali Imran (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-24648.txt.

Introduction

Omeke and co-authors design an energy-efficient federated anomaly detection system for underwater networks. Underwater acoustic links are slow, costly in energy, and often cannot reach the surface gateway. Their three-tier framework pairs sensors with nearby fog nodes in a feasibility-aware way. Model updates are compressed before sending, and fog nodes cooperate selectively, exchanging only when small clusters gain from larger neighbours. A physics-based acoustic model scores detection quality, energy use, and participation. In a 200-sensor test, only 48 percent could reach the gateway.

Catalog record

CREATORKenechi Omeke, Michael Mollel, Lei Zhang, Qammer H. Abbasi, Muhammad Ali Imran
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE58573
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.24648

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Energy-Efficient Hierarchical Federated Anomaly Detection for the Internet of Underwater Things via Selective Cooperative Aggregation Kenechi Omeke, Michael Mollel, Lei Zhang, Qammer H. Abbasi, Muhammad Ali Imran Abstract Anomaly detection is a core service in the Internet of Underwater Things, yet training accurate distributed models underwater is difficult because acoustic links are low-bandwidth, energy-intensive, and often unable to support direct sensor-to-surface communication. Standard flat federated learning therefore faces two coupled limitations in underwater deployments: expensive long-range transmissions and reduced participation when only a subset of sensors can reach the gateway. This paper proposes an energy-efficient hierarchical federated learning framework for underwater anomaly detection based on three components: feasibility-aware sensor-to-fog association, compressed model-update transmission, and selective cooperative aggregation among fog nodes. The proposed three-tier architecture localises most communication within short-range clusters while activating fog-to-fog exchange only when smaller clusters can benefit from nearby larger neighbours. A physics-grounded underwater acoustic model is used to evaluate detection quality, communication energy, and network participation jointly. In large synthetic deployments, only about 48% of sensors can directly reach the gateway in the 200-sensor case, whereas hierarchical learning preserves full participation through feasible fog paths. Selective cooperation matches the detection accuracy of always-on inter-fog exchange while reducing its energy by 31–33%, and compressed uploads reduce total energy by 71–95% in matched sensitivity tests. Experiments on three real benchmarks further show that low-overhead hierarchical methods remain competitive in detection quality, while flat federated learning defines the minimum-energy operating point. These results provide practical design guidance for underwater deployments operating under severe acoustic communication constraints. I Introduction The Internet of Underwater Things (IoUT) extends the terrestrial Internet of Things (IoT) into subsea environments, enabling persistent ocean observation, offshore infrastructure monitoring, environmental protection, and autonomous underwater operations [ 3 , 33 , 14 , 2 ] . As these deployments grow in scale and autonomy, they increasingly require in-network intelligence rather than simple data collection. Among the most important of these intelligent services is anomaly detection , which enables early discovery of sensor faults, abnormal environmental events, and malicious behaviour from distributed multivariate time series. Delivering such intelligence underwater is far more difficult than in terrestrial IoT. IoUT nodes communicate predominantly through acoustic links whose low bandwidth, long propagation delay, high bit-error rate, and severe attenuation make frequent long-range communication expensive and unreliable [ 40 , 50 ] . Combined with limited onboard energy and the practical difficulty of battery replacement or maintenance, these channel characteristics make raw-data centralisation unattractive for learning tasks. Federated learning (FL) is therefore a natural candidate for IoUT because it avoids transferring raw measurements and instead exchanges model updates [ 26 , 17 ] . However, conventional star-topology FL still assumes that every sensor can repeatedly communicate with a central coordinator. Underwater, this assumption is problematic for two reasons. First, direct sensor-to-surface transmissions are often the most energy-expensive links in the network. Second, some sensors may not be able to reach the surface gateway at all under realistic acoustic constraints. As a result, a flat FL system can appear highly efficient while training on only the directly reachable subset of the deployment. Hierarchical federated learning (HFL) offers a promising alternative by inserting fog or edge aggregators between the sensors and the global server [ 22 , 24 , 46 ] . In IoUT, this hierarchy is physically meaningful: clusters of seafloor sensors can upload to nearby mid-water aggregators, such as autonomous underwater vehicles (AUVs) or anchored relays, which then communicate aggregated models to a surface gateway. This architecture reduces the burden of long-range transmissions, but it introduces a second design challenge. If fog-to-fog model exchange is enabled indiscriminately, the resulting inter-fog traffic may erase a large part of the expected energy gain. The central problem addressed in this paper is therefore not simply whether HFL is better than flat FL, but how hierarchical cooperation should be designed for underwater learning under realistic acoustic constraints . A useful IoUT learning framework must answer three practical questions at once: which sensors can actually participate, how much communication energy can be reduced through compressed uplinks, and when cooperative aggregation among fog nodes is beneficial enough to justify its additional cost. Existing IoUT studies on FL do not address these questions jointly and generally stop short of a physics-grounded evaluation of learning quality, communication energy, and effective network participation [ 31 , 32 , 37 ] . s_{1} s_{2} s_{3} s_{4} s_{5} s_{6} Fog f_{1} (AUV) Fog f_{2} (AUV) Surface Gateway Cloud Association and cooperation rules cooperative aggregation water surface Deep (500–1000 m) Mid-water (100–400 m) Fig. 1: Proposed three-tier HFL architecture for IoUT with feasibility-aware association, compressed uplinks, and selective fog cooperation. In response, this paper proposes a three-tier hierarchical federated anomaly-detection framework with selective cooperative aggregation for IoUT, illustrated in Fig. 1 . The proposed framework combines feasibility-aware sensor-to-fog association, compressed model-update transmission, and selective fog-to-fog cooperation so that the communication architecture remains energy-efficient without sacrificing the ability to learn from the full underwater deployment. Equally importantly, the framework is constructed to evaluate the learning system in the way an IoUT designer actually needs to see it: not only in terms of detection accuracy and energy consumption, but also in terms of how much of the network can effectively participate in training. Its central contribution is to make these feasibility constraints explicit within a participation-aware evaluation framework, allowing the flat-versus-hierarchical design trade-off to be quantified under realistic IoUT conditions. The main contributions of the paper are as follows: 1. Participation-aware hierarchical IoUT learning framework. We formulate a three-tier underwater learning system (sensor \rightarrow fog \rightarrow surface) with a capped-source-level acoustic channel model, feasible communication graph, and explicit participation accounting, enabling realistic evaluation of who can train, at what communication cost, and through which underwater paths. 2. Selective cooperative aggregation with compressed uplinks. We introduce a lightweight inter-fog cooperation strategy that activates only when smaller fog clusters can benefit from nearby larger neighbours, and we combine it with compressed sensor uploads to make hierarchical training practical over acoustic links. 3. Comprehensive evaluation and IoUT design guidance. Through synthetic experiments and real-data evaluation on the Server Machine Dataset (SMD), Soil Moisture Active Passive (SMAP), and Mars Science Laboratory (MSL) benchmarks, we quantify the trade-offs among participation, detection accuracy, and energy consumption, and derive practical guidance for when flat FL is sufficient and when hierarchy becomes necessary to retain feasible participation. Organisation. Section II surveys related work. Section III presents the system and channel model. Section IV formalises the optimisation problem and decision space. Section V details the proposed hierarchical FL framework. Section VI presents implementation details and performance results, and Section VII concludes the paper. II Related Work This section reviews three intersecting threads: IoUT architectures and anomaly detection; federated and hierarchical FL; and communication- and energy-efficient FL. Adaptive orchestration, including RL-based approaches, is noted only briefly as background because the main paper studies transparent deterministic association and cooperation rules rather than learned controllers. It then synthesises the gaps that motivate the proposed hierarchical IoUT learning framework. II-A IoUT Architectures, Anomaly Detection, and Edge Intelligence IoUT extends underwater sensor networks and AUV systems toward persistent, data-driven subsea cyber-physical systems [ 3 , 33 , 14 ] . Acoustic

Read the complete text