Scholarlink

<p>The proliferation of Internet of Things (IoT) devices in critical infrastructure has created unprecedented cybersecurity challenges, particularly in resource-constrained environments where traditional security solutions prove inadequate. This paper presents a novel Adaptive Federated Edge Intelligence framework that combines lightweight deep learning with federated learning principles to enable real-time cyberthreat detection in IoT ecosystems. Our approach addresses the fundamental limitations of centralized security architectures by distributing intelligence across edge devices while preserving privacy and minimizing computational overhead. The proposed Adaptive Resource-Aware Federated Aggregation (ARFA) algorithm dynamically adjusts model complexity based on device capabilities, ensuring optimal performance across heterogeneous IoT networks. We introduce a lightweight convolutional neural network architecture specifically designed for resource-constrained environments, achieving 94.7% detection accuracy while consuming only 12.3 MB of memory and 45 milliseconds of processing time per threat analysis. The federated learning mechanism enables collaborative threat intelligence sharing without exposing sensitive local data, addressing critical privacy concerns in distributed IoT deployments. Experimental evaluation on three realistic IoT attack datasets demonstrates superior performance compared to existing centralized and edge-based approaches, with 23% faster response times and 41% lower energy consumption. The framework successfully adapts to evolving threat landscapes through continuous learning while maintaining strict resource constraints. These results establish the viability of adaptive federated edge intelligence as a scalable solution for next-generation IoT cybersecurity, offering significant implications for critical infrastructure protection and autonomous system security.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Adaptive Federated Edge Intelligence for Real-Time Cyberthreat Detection in Resource-Constrained IoT Environments: A Lightweight Deep Learning Approach

  • 期刊论文

  • 发表日期：

    2025年10月22日

查看全文
  • Milad Rahmati,
  • Nima Rahmati

摘要

The proliferation of Internet of Things (IoT) devices in critical infrastructure has created unprecedented cybersecurity challenges, particularly in resource-constrained environments where traditional security solutions prove inadequate. This paper presents a novel Adaptive Federated Edge Intelligence framework that combines lightweight deep learning with federated learning principles to enable real-time cyberthreat detection in IoT ecosystems. Our approach addresses the fundamental limitations of centralized security architectures by distributing intelligence across edge devices while preserving privacy and minimizing computational overhead. The proposed Adaptive Resource-Aware Federated Aggregation (ARFA) algorithm dynamically adjusts model complexity based on device capabilities, ensuring optimal performance across heterogeneous IoT networks. We introduce a lightweight convolutional neural network architecture specifically designed for resource-constrained environments, achieving 94.7% detection accuracy while consuming only 12.3 MB of memory and 45 milliseconds of processing time per threat analysis. The federated learning mechanism enables collaborative threat intelligence sharing without exposing sensitive local data, addressing critical privacy concerns in distributed IoT deployments. Experimental evaluation on three realistic IoT attack datasets demonstrates superior performance compared to existing centralized and edge-based approaches, with 23% faster response times and 41% lower energy consumption. The framework successfully adapts to evolving threat landscapes through continuous learning while maintaining strict resource constraints. These results establish the viability of adaptive federated edge intelligence as a scalable solution for next-generation IoT cybersecurity, offering significant implications for critical infrastructure protection and autonomous system security.