Scholarlink

<p>In the modern, rapidly evolving cybersecurity environment, smart homes face the constant challenge of protecting against both conventional and emerging zero-day threats. A smart home typically contains a significant number of interconnected IoT (Internet of Things) devices, which work together to automate and enhance daily living. These devices are connected via the internet or local networks and often communicate with each other via a central controller, like a smartphone app, hub, or router. New signature-based attacks are particularly difficult to detect due to limited prior information and poor detection rates. This paper introduces a modified federated learning based novel multi-stage intrusion detection system for real-time identification of attacks in a smart home ecosystem. Using the CICIDS2017 and CICIoT2023 datasets, a bio-inspired metaheuristic, Walrus Optimization Algorithm, is first applied to select the most informative features. The system then employs a three-stage detection process: lightweight machine learning models in the first stage for quick identification of known attacks, an ensemble learning framework in the second stage for deeper analysis, and an autoencoder-based deep learning module in the third stage to detect zero-day attacks misclassified earlier. Overall, integrated frameworks are periodically updated with their optimized hyperparameters using a modified federated learning framework that ensures privacy-preserving model aggregation across distributed IoT devices while minimizing communication overhead and adapting to dynamic network conditions. Experimental evaluations have been conducted based on six evaluation metrics: accuracy, precision, recall, F1-score, ROC curve, and execution time. Experimental results demonstrate that the proposed approach achieves over 99% accuracy with high precision, recall, and F1-score values and outperforms state-of-the-art techniques. A custom-built real-time testbed further demonstrates its scalability and effectiveness in handling live network traffic, making it a reliable solution for a modern smart home ecosystem.</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

An enhanced federated learning-based multistage IDS for detecting zero-day and known threats in smart home ecosystem

  • 期刊论文

  • 发表日期：

    2026年7月2日

查看全文
  • Rupali Banerjee,
  • Arpan Goswami,
  • Ratul Chowdhury,
  • Soumadip Mondal,
  • Khokan Mondal,
  • Tanmoy Chakraborty

摘要

In the modern, rapidly evolving cybersecurity environment, smart homes face the constant challenge of protecting against both conventional and emerging zero-day threats. A smart home typically contains a significant number of interconnected IoT (Internet of Things) devices, which work together to automate and enhance daily living. These devices are connected via the internet or local networks and often communicate with each other via a central controller, like a smartphone app, hub, or router. New signature-based attacks are particularly difficult to detect due to limited prior information and poor detection rates. This paper introduces a modified federated learning based novel multi-stage intrusion detection system for real-time identification of attacks in a smart home ecosystem. Using the CICIDS2017 and CICIoT2023 datasets, a bio-inspired metaheuristic, Walrus Optimization Algorithm, is first applied to select the most informative features. The system then employs a three-stage detection process: lightweight machine learning models in the first stage for quick identification of known attacks, an ensemble learning framework in the second stage for deeper analysis, and an autoencoder-based deep learning module in the third stage to detect zero-day attacks misclassified earlier. Overall, integrated frameworks are periodically updated with their optimized hyperparameters using a modified federated learning framework that ensures privacy-preserving model aggregation across distributed IoT devices while minimizing communication overhead and adapting to dynamic network conditions. Experimental evaluations have been conducted based on six evaluation metrics: accuracy, precision, recall, F1-score, ROC curve, and execution time. Experimental results demonstrate that the proposed approach achieves over 99% accuracy with high precision, recall, and F1-score values and outperforms state-of-the-art techniques. A custom-built real-time testbed further demonstrates its scalability and effectiveness in handling live network traffic, making it a reliable solution for a modern smart home ecosystem.