Scholarlink

<p>The rapid expansion of Internet of Things (IoT) networks has exposed critical vulnerabilities, necessitating advanced security solutions capable of real-time threat detection and mitigation. This study aims to develop and evaluate optimized AI- and deep learning-based security frameworks for IoT environments, addressing limitations in existing methods, such as low accuracy in anomaly detection, delayed response to emerging cyberattacks, and inefficiency in resource-constrained devices. The proposed Online Sequential Ensemble Learning approach integrates ensemble learning, feature selection, and adaptive optimization to enhance model performance and computational efficiency. Experiments were conducted on multiple benchmark datasets—including CICDoS2019, NSL-KDD, UNSW-NB15, ToN-IoT, and LATAM-DDoS-IoT—demonstrating high efficacy, with accuracy ranging from 91.9 to 100% and error rates as low as 0.1%, outperforming traditional ML and DL techniques. These results confirm that the improvements stem from both the optimized learning framework and the underlying model design. The study helps bridge gaps in IoT security research by providing a scalable, energy-efficient, and robust detection system. In conclusion, the proposed method provides a reliable reference for designing secure IoT networks, highlighting the potential to integrate optimized cryptography and AI-driven adaptive mechanisms in future large-scale deployments.</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

A detailed comparative study on cybersecurity in IoT environment

  • 期刊论文

  • 发表日期：

    2026年1月14日

查看全文
  • Urvashi Sangwan,
  • Rajender Singh Chhillar

摘要

The rapid expansion of Internet of Things (IoT) networks has exposed critical vulnerabilities, necessitating advanced security solutions capable of real-time threat detection and mitigation. This study aims to develop and evaluate optimized AI- and deep learning-based security frameworks for IoT environments, addressing limitations in existing methods, such as low accuracy in anomaly detection, delayed response to emerging cyberattacks, and inefficiency in resource-constrained devices. The proposed Online Sequential Ensemble Learning approach integrates ensemble learning, feature selection, and adaptive optimization to enhance model performance and computational efficiency. Experiments were conducted on multiple benchmark datasets—including CICDoS2019, NSL-KDD, UNSW-NB15, ToN-IoT, and LATAM-DDoS-IoT—demonstrating high efficacy, with accuracy ranging from 91.9 to 100% and error rates as low as 0.1%, outperforming traditional ML and DL techniques. These results confirm that the improvements stem from both the optimized learning framework and the underlying model design. The study helps bridge gaps in IoT security research by providing a scalable, energy-efficient, and robust detection system. In conclusion, the proposed method provides a reliable reference for designing secure IoT networks, highlighting the potential to integrate optimized cryptography and AI-driven adaptive mechanisms in future large-scale deployments.