Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{187353,
        author = {Devendra Bodkhe and Vaidehi Patil and Swaleha Deshmukh and Ujjawal Pathak},
        title = {Introduction and Motivation: Contextualizing the AIoT Grand Challenge},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {12},
        number = {6},
        pages = {4306-4314},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=187353},
        abstract = {The rapid expansion of the AIoT market is driving critical resource-constrained applications, such as sustainable utility monitoring in smart cities. Deploying sophisticated AI models at the edge (TinyML) faces a triple challenge: severe power limitations, vulnerability to adversarial attacks, and privacy risks inherent in Federated Learning (FL) aggregation. To resolve this, we propose PRAM-SU (Privacy-Preserving and Adversarial Robust TinyML Framework for Sustainable Edge Utility Monitoring), a novel architecture optimized for NPU-accelerated microcontrollers. PRAM-SU introduces a Deep Reinforcement Learning (DRL) agent that guides model pruning to selectively target and reduce the energy consumption of high-power convolutional layers, significantly enhancing efficiency. Furthermore, we implement a lightweight, online adversarial training module to ensure system robustness, resulting in minimal performance degradation when subjected to state-of-the-art attacks (e.g., FGSM, PGD). Critical to data protection, we establish a quantitative privacy evaluation using metrics like Fréchet Inception Distance (FID) and Generative Model Inversion (GMI) to validate secure FL model aggregation. Empirical validation using real-world traffic data demonstrates that PRAM-SU provides an end-to-end, resource-efficient, and secure solution crucial for next-generation sustainable Edge AI deployment.},
        keywords = {TinyML, Edge AI, Federated Learning (FL), Adversarial Robustness, Deep Reinforcement Learning (DRL), Quantitative Privacy, Smart City, Sustainable Monitoring, Neural Processing Unit (NPU).},
        month = {November},
        }