Scholarlink

<p>The advancement of 5&#xa0;G/6&#xa0;G networks and the internet of things has unlocked a plethora of connectivity and application prospects, simultaneously escalating risks to network infrastructures. The surge in connected devices and the immense volumes of data create a pressing need for timely information processing, prompting a shift from centralized solutions to edge computing for rapid response to cyber threats. However, due to the constraints of edge devices, attack detection solutions consistently encounter significant issues such as precision, resource allocation, and stability. To tackle these challenges head-on, we introduce a novel attack detection model aimed at optimizing efficiency and improving accuracy while clarifying machine learning approaches’ transparency. This proposed model combines five machine learning algorithms that act as experts with novel developed feature selection methods in explainable AI (XAI) as SHAP and LIME Methods. The multiple expert systems are coordinated in the decision-making framework, encompassing the entire decision to fit practical conditions. We found that our model does better than state-of-the-art studies in rate accuracy, decision-making latency, and network resource optimization when we used the CiCIoT dataset to validate its performance.</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 hybrid intrusion detection system model integrated explainable AI and multi expert systems to adapt edge computing

  • 期刊论文

  • 发表日期：

    2025年9月3日

查看全文
  • Trong-Minh Hoang,
  • Van-Nhan Nguyen,
  • Trang-Linh Le Thi,
  • Minh-Hoang Nguyen,
  • Nam-Hoang Nguyen

摘要

The advancement of 5 G/6 G networks and the internet of things has unlocked a plethora of connectivity and application prospects, simultaneously escalating risks to network infrastructures. The surge in connected devices and the immense volumes of data create a pressing need for timely information processing, prompting a shift from centralized solutions to edge computing for rapid response to cyber threats. However, due to the constraints of edge devices, attack detection solutions consistently encounter significant issues such as precision, resource allocation, and stability. To tackle these challenges head-on, we introduce a novel attack detection model aimed at optimizing efficiency and improving accuracy while clarifying machine learning approaches’ transparency. This proposed model combines five machine learning algorithms that act as experts with novel developed feature selection methods in explainable AI (XAI) as SHAP and LIME Methods. The multiple expert systems are coordinated in the decision-making framework, encompassing the entire decision to fit practical conditions. We found that our model does better than state-of-the-art studies in rate accuracy, decision-making latency, and network resource optimization when we used the CiCIoT dataset to validate its performance.