Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LLMs meet Federated Learning for Scalable and Secure IoT Management

Yazan Otoum, Arghavan Asad, Amiya Nayak

TL;DR

This work tackles the need for scalable, privacy-preserving IoT management by combining large language models with federated learning. It introduces a gradient-sensing federated strategy (GSFS) within a hybrid edge-cloud architecture to adaptively regulate client participation and updates under real-time network conditions. Across IoT-23 tasks, GSFS demonstrates higher central and client accuracy and notably lower latency and better energy efficiency compared with FedAvg and FedOpt, illustrating the practical value of LLM-powered FL in large-scale IoT systems. The framework enables secure, adaptive, and intelligent IoT management with reduced communication overhead and improved robustness in distributed, heterogeneous environments.

Abstract

The rapid expansion of IoT ecosystems introduces severe challenges in scalability, security, and real-time decision-making. Traditional centralized architectures struggle with latency, privacy concerns, and excessive resource consumption, making them unsuitable for modern large-scale IoT deployments. This paper presents a novel Federated Learning-driven Large Language Model (FL-LLM) framework, designed to enhance IoT system intelligence while ensuring data privacy and computational efficiency. The framework integrates Generative IoT (GIoT) models with a Gradient Sensing Federated Strategy (GSFS), dynamically optimizing model updates based on real-time network conditions. By leveraging a hybrid edge-cloud processing architecture, our approach balances intelligence, scalability, and security in distributed IoT environments. Evaluations on the IoT-23 dataset demonstrate that our framework improves model accuracy, reduces response latency, and enhances energy efficiency, outperforming traditional FL techniques (i.e., FedAvg, FedOpt). These findings highlight the potential of integrating LLM-powered federated learning into large-scale IoT ecosystems, paving the way for more secure, scalable, and adaptive IoT management solutions.

LLMs meet Federated Learning for Scalable and Secure IoT Management

TL;DR

This work tackles the need for scalable, privacy-preserving IoT management by combining large language models with federated learning. It introduces a gradient-sensing federated strategy (GSFS) within a hybrid edge-cloud architecture to adaptively regulate client participation and updates under real-time network conditions. Across IoT-23 tasks, GSFS demonstrates higher central and client accuracy and notably lower latency and better energy efficiency compared with FedAvg and FedOpt, illustrating the practical value of LLM-powered FL in large-scale IoT systems. The framework enables secure, adaptive, and intelligent IoT management with reduced communication overhead and improved robustness in distributed, heterogeneous environments.

Abstract

The rapid expansion of IoT ecosystems introduces severe challenges in scalability, security, and real-time decision-making. Traditional centralized architectures struggle with latency, privacy concerns, and excessive resource consumption, making them unsuitable for modern large-scale IoT deployments. This paper presents a novel Federated Learning-driven Large Language Model (FL-LLM) framework, designed to enhance IoT system intelligence while ensuring data privacy and computational efficiency. The framework integrates Generative IoT (GIoT) models with a Gradient Sensing Federated Strategy (GSFS), dynamically optimizing model updates based on real-time network conditions. By leveraging a hybrid edge-cloud processing architecture, our approach balances intelligence, scalability, and security in distributed IoT environments. Evaluations on the IoT-23 dataset demonstrate that our framework improves model accuracy, reduces response latency, and enhances energy efficiency, outperforming traditional FL techniques (i.e., FedAvg, FedOpt). These findings highlight the potential of integrating LLM-powered federated learning into large-scale IoT ecosystems, paving the way for more secure, scalable, and adaptive IoT management solutions.

Paper Structure

This paper contains 21 sections, 14 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Proposed Framework
  4. Architecture Overview
  5. Implementation Details
  6. Classical Federated Learning Baselines
  7. FedAvg.
  8. FedOpt.
  9. Client Side
  10. Performance-Based Trigger.
  11. Adaptive Threshold Based on Gradient Norms.
  12. Server Side
  13. Caching and Aggregation Trigger.
  14. Upload and Broadcast Design
  15. Consistent Model Broadcasting.
  16. ...and 6 more sections

Figures (2)

  • Figure 1: Architecture of our proposed IoT management system integrating edge and cloud computing with federated learning. The framework includes key components such as local LLM training at the edge, a federated learning coordinator for global model updates, a communication module for secure and efficient data transfer, etc..
  • Figure 2: Loss curves for the OPT-350M and OPT-1.3b models over training epochs.