Scholarlink

Hackers have more avenues for attack due to the abundance of Internet of Things (IoT) devices, which are often weak and insecure. This research presents a new integrated deep learning model combining Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM) networks, and Gated Recurrent Units (GRUs) for improved botnet detection in IoT networks. The model leverages LSTMs for time-dependent patterns, GRUs for short-term patterns with efficiency, and CNNs for spatial feature extraction. Although the integrated architecture enhances spatiotemporal traffic comprehension, it complicates understanding. Using the UNSW-NB15 and BotNet-IoT datasets, tests showed exceptional detection capabilities. The model achieved 99.20% accuracy, 99.12% ROC-AUC, and 98.30% F1 score on UNSW-NB15, and 99.99% accuracy, 99.99% ROC-AUC, and 99.98% F1 score on Bot-IoT. The CNN-LSTM-GRU model excels in dynamic IoT environments, detecting complex botnet activities better than standard IDS and basic deep learning techniques. This study highlights the need for hybrid learning methods to combat emerging cyber threats and enhance IoT security.

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

Botnet Detection in IoT Networks Using Hybrid Deep Learning CNN, LSTM, and GRU

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • A. S. S. Adil,
  • Haider K. Hoomod

摘要

Hackers have more avenues for attack due to the abundance of Internet of Things (IoT) devices, which are often weak and insecure. This research presents a new integrated deep learning model combining Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM) networks, and Gated Recurrent Units (GRUs) for improved botnet detection in IoT networks. The model leverages LSTMs for time-dependent patterns, GRUs for short-term patterns with efficiency, and CNNs for spatial feature extraction. Although the integrated architecture enhances spatiotemporal traffic comprehension, it complicates understanding. Using the UNSW-NB15 and BotNet-IoT datasets, tests showed exceptional detection capabilities. The model achieved 99.20% accuracy, 99.12% ROC-AUC, and 98.30% F1 score on UNSW-NB15, and 99.99% accuracy, 99.99% ROC-AUC, and 99.98% F1 score on Bot-IoT. The CNN-LSTM-GRU model excels in dynamic IoT environments, detecting complex botnet activities better than standard IDS and basic deep learning techniques. This study highlights the need for hybrid learning methods to combat emerging cyber threats and enhance IoT security.