Pavan Sai Ramarao Maddali, Vijay Kumar Damera, Ratna Kumar Prathipati
Edge IoT deployments must evaluate the trustworthiness of peer devices continuously, locally, and under resource constraints that static scoring models cannot accommodate—particularly when adversaries adapt their mimicry strategies in response to observed detection outcomes. This study presents the Adaptive Dynamic Trust Evaluation (ADTE) framework, with three core novelties that address the distinct failure modes of prior research. First, an edge-masked heterogeneous attention (EMHA) operator suppresses adversarially injected graph edges through a learnable perturbation-aware gate, for which we prove a certified robustness radius under ℓ ∞ -bounded node-feature perturbation and quantify mask convergence under PGD-augmented training. We additionally demonstrate empirical—though not formally certified—robustness against gradient-based edge-injection attacks and explicitly define the gap between the analytic certificate (node features) and the main graph-poisoning threat model (edge structure). Second, a non-stationary MDP formulation models the adversary’s policy as a time-varying component of the transition kernel; we derive a regret bound of O (T Δ T) that holds under both piece-wise stationary and continuously drifting attack distributions by applying a sliding-window restart mechanism, proving that this bound extends to partially observable adversarial policies via a POMDP belief-state reduction. Third, a continual PPO agent augmented with elastic weight consolidation (EWC) prevents catastrophic forgetting over long deployment horizons; we demonstrate stable policy performance across 5000-episode simulations spanning ≈ 3.47 continuous deployment-days at 60 s evaluation windows (equivalent to 300,000 s of gateway operation). Unlike prior trust evaluation frameworks that address, at most, two of the three sub-problems (representation, decision, and adaptation), ADTE simultaneously targets all three, thus delivering certified robustness guarantees that are absent from GNN-based approaches, non-stationary regret bounds missing from RL-based methods, and continual learning stability unavailable in either category. Evaluated on N-BaIoT, MedBIoT, and UNSW-NB15 against eight baselines, ADTE achieves 95.3% trust classification accuracy, 41.8% FPR reduction over FedTrust, and sub-12 m edge latency at 1.8 W incremental power. The 41.8% FPR reduction over FedTrust is measured under equivalent single-gateway operating conditions. These results are validated under gradient-based adaptive graph poisoning, asynchronous decentralised gateway operation, and topology churn up to 1,200 devices—conditions no prior trust evaluation framework has simultaneously addressed.
Targeted monitoring of large natural areas through the Internet of Things remains a significant challenge for wildlife conservation. While IoT systems can detect wildlife poaching, the scale of conservation areas makes blanket coverage infeasible. Therefore, identifying key hotspots for IoT deployment is essential. Existing geospatial and clustering methods assist in hotspot identification, but no prior work provided a complete cold-start pipeline from community memory to a community-derived hotspot map when no prior incident records exist. This study presents a five-stage geospatial data pipeline for the 2,000-acre, ecologically diverse campus of the University of Hyderabad India. The stages are: (1) campus digitization using OpenStreetMap and QGIS; (2) feature extraction from Shuttle Radar Topography Mission, Copernicus, WorldPop, and PROBA-V, yielding a spatial dataset of 13,000 points with 40 features each; (3) computation of distance vectors to all digitized natural and infrastructural elements; (4) community labeling (11–12 respondents) of K-Means clusters using a ternary scheme; and (5) label propagation and hotspot map generation. The descriptive analysis conducted in this study showed that community-identified poaching areas exhibit ecologically coherent spatial patterns compared to areas labeled as non-poaching and ML classification confirmed the labels are spatially learnable, with KNN model reaching a weighted F1 of 0.96. The pipeline extends prior art into a substantially more challenging context: no historical records, responses from non-expert community members, and an explicit uncertainty representation. The resulting hotspot map helped the development and deployment of a fog-based, energy-efficient IoT monitoring system, which is also discussed in the paper. The pipeline is applicable to any conservation area where official records are absent, and community knowledge is the only available ground-truth resource.
Zhiyu Wang, Nilotpal Kapri, Luiz Fernando Bittencourt, Rajkumar Buyya
The rapid proliferation of Internet of Things (IoT) devices across domains such as smart transportation, healthcare, and smart city infrastructure has intensified the demand for low-latency, energy-efficient, and scalable computing paradigms. While cloud computing has traditionally served as the backbone for IoT data processing, its inherent limitations have catalyzed the emergence of edge/fog computing, forming a distributed edge-cloud continuum. This review examines emerging trends in Artificial Intelligence (AI)-driven resource management within this continuum, with a focus on three directions: (1) the transition from centralized to distributed and collaborative intelligence, (2) cross-domain adaptation and knowledge transfer for heterogeneous IoT applications, and (3) the nascent integration of foundation models into edge environments. We ground our discussion in three complementary case studies: smart transportation as a representative vertical domain, cross-domain heterogeneous IoT application scheduling as a horizontal perspective, and industrial transferred arc plasma monitoring as an emerging industrial IoT scenario, and conclude with forward-looking research directions, including quantum-enhanced edge optimization, edge-native continual learning, digital twin-driven resource orchestration, and neuromorphic computing for ultra-low-power edge AI.
Somnath Bera, Tanushree Dey, Anirbit Sengupta, Anwesha Mukherjee, Debashis De
Soil water management is one of the significant domains in smart irrigation. This paper proposes and implements an IoT and edge computing-based soil water management system for sustainable irrigation using ensemble learning. An ultralow power sensor node is used to collect real-time soil moisture and soil temperature data. The edge device performs raw data processing and then sends the pre-processed data to the cloud server after adding Gaussian noise to it for data privacy protection. Ensemble learning models are used for analyzing the data inside the cloud. The results are compared to single classifier-based machine learning models and deep learning models. The results demonstrate that the ensemble learning models outperform the single classifier-based machine learning models in terms of prediction accuracy, and the deep learning models in terms of latency and energy consumption. We also perform soil temperature and soil moisture prediction using ensemble learning to deal with the scenario, where either the soil moisture or soil temperature sensor is available, and the results show that the ensemble learning models outperform the baselines. We also analyze the proposed framework’s performance for other datasets to further demonstrate its efficacy. The results show that the proposed irrigation decision making framework has ≥99% prediction accuracy for our own dataset and publicly available datasets. Finally, an Android App is designed using which the user can access the data remotely, and can control the irrigation decision making process based on the results.
Recent advances in artificial intelligence (AI) have positioned large language models (LLMs) as potentially transformative tools for education. However, their adoption in primary schools raises significant concerns regarding student data privacy, institutional governance, and the affordability of AI infrastructure. This paper presents a conceptual analysis of locally hosted small language models (SLMs) as an alternative to cloud-dependent AI systems in primary education. The study examines developmental, institutional, and regulatory factors relevant to primary school settings and proposes a school-based architecture integrating low-cost hardware and open-source software frameworks operating entirely within local school networks. The proposed architecture demonstrates how locally deployed SLMs in primary schools can support educational AI applications while reducing dependence on external cloud providers. By maintaining data processing and storage within school networks, the framework may strengthen data sovereignty, improve institutional control over AI operations, and enhance accessibility for resource-constrained schools through lower infrastructure costs. The findings suggest that, within primary education, institutional controllability, privacy protection, and equitable access may be more valuable objectives than maximising model capability. This conceptual framework contributes to the emerging literature on educational AI governance by outlining a sociotechnical pathway that prioritises child data protection, institutional autonomy, and practical deployment feasibility while supporting the responsible integration of AI into educational environments.
