Scholarlink

<p>With the increasing expansion of Internet of Things (IoT) networks and the rising number of Internet-connected devices, securing these networks from cyber attacks is urgently needed. A major risk to IoT services is service disruption caused by Denial of Service (DoS) and Distributed Denial of Service (DDoS) attacks. This downtime is extremely dangerous in vital sectors. Many researchers have worked to enhance security by developing Intrusion Detection Systems (IDSs) to detect attacks on these networks. This research proposed a parallel hybrid deep learning model. It integrates Artificial Neural Network (ANN), Long Short-Term Memory (LSTM) neural networks, and Gated Recurrent Units (GRU) to detect DoS and DDoS attacks in IoT network traffic data. The Synthetic Minority Oversampling Technique (SMOTE) addresses the training data imbalance. The proposed model was applied to the CICIoT2023 dataset. Precision, Recall, F1-Score, and Accuracy criteria estimated the model’s performance. The outcomes revealed that our model achieved a Precision of 99.95%, an F1-Score of 99.95%, a Recall of 99.95% and an Accuracy of 99.95% with a low false positive rate.</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 parallel hybrid deep learning model for DoS and DDoS detection in IoT environments

  • 期刊论文

  • 发表日期：

    2026年6月15日

查看全文
  • Maha Alshehri,
  • Mrim Alnfiai

摘要

With the increasing expansion of Internet of Things (IoT) networks and the rising number of Internet-connected devices, securing these networks from cyber attacks is urgently needed. A major risk to IoT services is service disruption caused by Denial of Service (DoS) and Distributed Denial of Service (DDoS) attacks. This downtime is extremely dangerous in vital sectors. Many researchers have worked to enhance security by developing Intrusion Detection Systems (IDSs) to detect attacks on these networks. This research proposed a parallel hybrid deep learning model. It integrates Artificial Neural Network (ANN), Long Short-Term Memory (LSTM) neural networks, and Gated Recurrent Units (GRU) to detect DoS and DDoS attacks in IoT network traffic data. The Synthetic Minority Oversampling Technique (SMOTE) addresses the training data imbalance. The proposed model was applied to the CICIoT2023 dataset. Precision, Recall, F1-Score, and Accuracy criteria estimated the model’s performance. The outcomes revealed that our model achieved a Precision of 99.95%, an F1-Score of 99.95%, a Recall of 99.95% and an Accuracy of 99.95% with a low false positive rate.