Scholarlink

<p>The rapid integration of the Internet of Things (IoT) into healthcare systems has significantly increased exposure to cyberattacks, making accurate and low-latency intrusion detection a critical challenge. This paper proposes DQ-PAA, an autonomous intrusion detection architecture that combines Horse Herd Optimization (HHO)-based feature selection with a dual cellular Q-Learning decision framework. In the proposed method, HHO is employed to select the most informative network features, while a deep learning-based anomaly detector identifies suspicious traffic patterns. These detected anomalies are subsequently analyzed by a dual Q-Learning mechanism to optimize response decisions and reduce false alarms. The proposed DQ-PAA architecture is evaluated on benchmark intrusion detection datasets widely used in IoT security research, and its performance is compared against Bi-LSTM, LSTM, GMDH, standard Q-Learning, MLP, and NN models. Experimental results demonstrate that DQ-PAA achieves an average improvement of 5.73% in accuracy, 7.37% in precision, 6.78% in recall, and 7% in F-measure, while reducing training and prediction times by 9.63&#xa0;s and 10.71&#xa0;s, respectively, compared to the baseline methods. These results confirm that DQ-PAA provides a reliable and computationally efficient solution for intrusion detection in IoT-based healthcare environments.</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

DQ-PAA: an efficient autonomous architecture based on dual cellular Q-LEARNING for intrusion detection in the healthcare system in the internet of things context

  • 期刊论文

  • 发表日期：

    2026年6月29日

查看全文
  • Milad Tajik,
  • Hassan Shakeri,
  • Ali Akbar Neghabi,
  • Yasser Elmi Sola,
  • Mohammad Hossein Moattar

摘要

The rapid integration of the Internet of Things (IoT) into healthcare systems has significantly increased exposure to cyberattacks, making accurate and low-latency intrusion detection a critical challenge. This paper proposes DQ-PAA, an autonomous intrusion detection architecture that combines Horse Herd Optimization (HHO)-based feature selection with a dual cellular Q-Learning decision framework. In the proposed method, HHO is employed to select the most informative network features, while a deep learning-based anomaly detector identifies suspicious traffic patterns. These detected anomalies are subsequently analyzed by a dual Q-Learning mechanism to optimize response decisions and reduce false alarms. The proposed DQ-PAA architecture is evaluated on benchmark intrusion detection datasets widely used in IoT security research, and its performance is compared against Bi-LSTM, LSTM, GMDH, standard Q-Learning, MLP, and NN models. Experimental results demonstrate that DQ-PAA achieves an average improvement of 5.73% in accuracy, 7.37% in precision, 6.78% in recall, and 7% in F-measure, while reducing training and prediction times by 9.63 s and 10.71 s, respectively, compared to the baseline methods. These results confirm that DQ-PAA provides a reliable and computationally efficient solution for intrusion detection in IoT-based healthcare environments.