Faculty of Applied Science

Federated Learning for Detecting Anomalies in IoT-Driven Smart Home Systems

Wed, 01 Jan 2025 00:00:00 GMT

Federated Learning for Detecting Anomalies in IoT-Driven Smart Home Systems Thevarajan, J.; Ravi, V.; Kirushanth;, S.; Ganesalingam, V.; Ahmadon, M. A.; Che Lah, N. S. B. The rapid proliferation of Internet of Things (IoT) devices in smart home environments has amplified significant cybersecurity challenges due to their outdated firmware, limited computational resources, and weak protection mechanisms. Traditional centralized anomaly detection approaches are ineffective in these environments owing to single points of failure, privacy risks, and limited scalability. To address these challenges, this study proposes a novel federated learning framework that integrates a lightweight hybrid architecture combining 1D Convolutional Neural Networks (1D-CNN) with shallow autoencoders to support both supervised and unsupervised anomaly detection. The supervised component enables robust binary classification of known attack patterns, while the unsupervised reconstruction mechanism enhances detection of previously unseen anomalies, thus balancing accuracy with adaptability. Distinct from prior approaches, the proposed framework employs dynamic quantization within the federated learning process, enabling substantial reductions in resource consumption without compromising detection performance. Experiments conducted on the TON-IoT dataset, with SMOTE-based class balancing and PCA-driven dimensionality reduction, under a realistic non-IID distribution across three federated clients, demonstrate that the proposed model achieves 97.39% test accuracy. This model is designed for deployment on resource-constrained devices, employing post-training dynamic range quantization that reduces the model size by 81.5% (42.60 KB to 7.87 KB) without compromising accuracy. The experiment performed using Raspberry Pi 4 confirmed sub-millisecond inference and real-time responsiveness, validating the practical viability of the framework for efficient anomaly detection in smart IoT systems.

Design and Development of Weigh-In-Motion Using Vehicular Telematics

Wed, 01 Jan 2020 00:00:00 GMT

Design and Development of Weigh-In-Motion Using Vehicular Telematics Kirushanth, S.; Kabaso, B. Identifying overloaded vehicles on a highway is essential for the safety of vehicles on the road as well as for the performance monitoring of highway infrastructure and planning. Traffic enforcement uses various weigh-in-motion (WIM) methods. Since Vehicular Telematics (VT) is favoured in the transport industry, using it for building a new WIM system to infer the payload of a vehicle at any road segment would be beneficial for the transport industry. This paper presents the effort taken to use VT data from onboard diagnostics modules and smartphones to infer the payload of a vehicle. The experiment done to find the correlation between VT data and the payload of a vehicle is discussed. Feature engineering was done; nine different settings were tested to find the best regression model. A multiple nonlinear regression model produced significant a p value of 6.322e-08 and an R-squared value of 0.8736. Results support the notion of using the VT data for nonintrusive measurement of the weight of a vehicle in motion.

Preliminary Assessment of Seagrass Diversity and Meadow Structure in the Kuchchaveli Coastal Region, Trincomalee, Sri Lanka

Thu, 01 Jan 2026 00:00:00 GMT

Preliminary Assessment of Seagrass Diversity and Meadow Structure in the Kuchchaveli Coastal Region, Trincomalee, Sri Lanka Keerthanaram, T.; Arjunan, K.; Priyanchanaa, S. Seagrass meadows are ecologically important shallow coastal ecosystems that support marine biodiversity, stabilize sediments, enhance coastal productivity, and provide essential nursery habitats for fish and invertebrates. However, limited information is available on the species composition and spatial distribution of seagrass communities along the eastern coast of Sri Lanka. This study presents a preliminary field-based assessment of seagrass cover, species diversity, and meadow structure in the Kuchchaveli coastal region, Trincomalee. Five line-intercept transects were established perpendicular to the shoreline at 250 m intervals, with each transect extending 50 m seaward. Quadrat photographs were used to estimate percentage cover. The overall mean seagrass cover was 46.8 ± 12.4%. Among the recorded species, Oceana serrulata (formerly Cymodocea serrulata) had the highest mean cover (21.6 ± 7.8%), followed by Cymodocea rotundata (16.4 ± 6.2%) and Enhalus acoroides (8.8 ± 4.1%). Based on relative species-cover proportions, the Shannon-Wiener diversity index (H′) was 1.04, and Simpson’s diversity index (1 − D) was 0.63. Species co-occurrence and variation in meadow form indicate spatially heterogeneous habitat conditions, although the roles of sediment type, depth, and hydrodynamics require further investigation. This study establishes baseline ecological information for a poorly studied seagrass habitat on Sri Lanka’s eastern coast. It provides a scientific foundation for long-term monitoring, habitat restoration, and marine spatial planning.

Fishermen’s Perceptions of Dugong (Dugong dugon) Conservation and Seagrass Ecosystems in the Northern Coastal Region of Sri Lanka

Thu, 01 Jan 2026 00:00:00 GMT

Fishermen’s Perceptions of Dugong (Dugong dugon) Conservation and Seagrass Ecosystems in the Northern Coastal Region of Sri Lanka Keerthanaram, T.; Priyanchanaa, S. The dugong (Dugong dugon) is a vulnerable marine mammal that depends primarily on shallow coastal seagrass meadows for food. Although dugongs historically occurred throughout the Gulf of Mannar and Palk Bay, baseline population data for northern Sri Lanka remain limited amid concerns about habitat degradation, incidental capture, hunting, coastal development, and vessel traffic. This study investigated fishermen’s perceptions and local ecological knowledge of dugongs and seagrass ecosystems along the northern coastal belt from Mannar Island to Jaffna Lagoon. A structured questionnaire was administered to respondents from coastal fishing communities using systematic spatial sampling based on 5 km × 5 km coastal grid cells (n = 60). Most respondents were familiar with dugongs (96%), 92% correctly identified the species, and 48% reported observing dugongs while fishing. Nearly three-quarters (72%) perceived that dugong abundance had declined over time. Ecological knowledge of seagrass was high: all respondents recognized seagrass beds, and 95% identified their importance as dugong feeding habitat. Accidental entanglement in fishing gear was the most frequently identified threat (78%), followed by habitat degradation (61%) and disturbance from motorized boat traffic (44%). The conservation-attitude scale showed strong internal reliability (Cronbach’s α = 0.84), with a mean score of 4.1 ± 0.63. Dugong knowledge was positively correlated with conservation attitudes (r = 0.53, p < 0.01) and was the strongest predictor in the regression analysis (p < 0.01). These findings demonstrate that fishermen possess valuable local ecological knowledge that can complement scientific monitoring and support participatory conservation planning and policy development for dugongs and seagrass habitats in northern Sri Lanka.