Scholarlink

<p>The Internet of Things (IoT) has revolutionized modern technology by inter- connecting numerous devices and systems, enabling innovative applications across various domains. However, the widespread adoption of IoT has also introduced significant cybersecurity challenges, necessitating advanced Network Intrusion Detection Systems (NIDS) to counter evolving threats. These challenges are further compounded by resource constraints, high-dimensional data, and computational overhead, making efficient NIDS crucial for robust security. To address these challenges, we propose a novel feature selection optimization technique—Transient Crow Search Optimization (TCSO)—which effectively identifies an optimized feature subset, enhancing detection accuracy and computational efficiency. Additionally, we incorporate a Knowledge Distillation (KD) approach, employing a two-phase teacher-student model to transfer knowledge to a lightweight model capable of accurately detecting and classifying various attacks. Extensive evaluations on benchmark IoT datasets demonstrate the robustness and superior performance of the proposed method, achieving 99% accuracy, low False Alarm Rates (FAR), and minimal computational time— 9.7, 9.9, and 21&#xa0;s for the NSL-KDD, ToN-IoT, and BoT-IoT datasets, respectively. The method significantly outperforms existing approaches and can detect attacks in just 9&#xa0;s, making it exceptionally fast. This research con- tributes to the development of secure and resilient IoT systems, advancing the state-of-the-art in Network Intrusion Detection Systems (NIDS). </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

TCSO-KD: a novel technique for optimal feature selection and knowledge distillation for efficient IoT intrusion detection systems

  • 期刊论文

  • 发表日期：

    2025年9月4日

查看全文
  • P. Akanksha,
  • S. Manohar Naik

摘要

The Internet of Things (IoT) has revolutionized modern technology by inter- connecting numerous devices and systems, enabling innovative applications across various domains. However, the widespread adoption of IoT has also introduced significant cybersecurity challenges, necessitating advanced Network Intrusion Detection Systems (NIDS) to counter evolving threats. These challenges are further compounded by resource constraints, high-dimensional data, and computational overhead, making efficient NIDS crucial for robust security. To address these challenges, we propose a novel feature selection optimization technique—Transient Crow Search Optimization (TCSO)—which effectively identifies an optimized feature subset, enhancing detection accuracy and computational efficiency. Additionally, we incorporate a Knowledge Distillation (KD) approach, employing a two-phase teacher-student model to transfer knowledge to a lightweight model capable of accurately detecting and classifying various attacks. Extensive evaluations on benchmark IoT datasets demonstrate the robustness and superior performance of the proposed method, achieving 99% accuracy, low False Alarm Rates (FAR), and minimal computational time— 9.7, 9.9, and 21 s for the NSL-KDD, ToN-IoT, and BoT-IoT datasets, respectively. The method significantly outperforms existing approaches and can detect attacks in just 9 s, making it exceptionally fast. This research con- tributes to the development of secure and resilient IoT systems, advancing the state-of-the-art in Network Intrusion Detection Systems (NIDS).