Scholarlink

Using a history book approach, this paper provides a state-of-the-art study in terms of deep and ensemble learning models to be exploited for IIoT network intrusion detection. We evaluate the performance of diverse models on a diverse dataset (BCCC-cPacket-Cloud-DDoS-2024) with diverse samples, attack types, and network traffic patterns, including Artificial Neural Network (ANN), Long Short-Term Memory (LSTM), XGBoost, and LightGBM. Model evaluation is performed over several metrics like accuracy, precision, recall, F1 score, ROC-AUC. The findings suggest that ensemble learning models, which includes XGBoost and LightGBM, provide better accuracy and robustness than other models, which makes them suitable for the real-time detection tasks. Deep learning algorithms perform well when dealing with complex patterns in the data, but need more computational power. It also offers fundamental insights to help in selecting optimal IDS model(s) with respect to performance and resource constraints for real-time and offline detection.

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

Advanced Intrusion Detection Systems: A Comparative Study of Deep Learning and Ensemble Learning Models for IIoT Networks

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Lahcen Idouglid,
  • Said Tkatek,
  • Khalid Elfayq

摘要

Using a history book approach, this paper provides a state-of-the-art study in terms of deep and ensemble learning models to be exploited for IIoT network intrusion detection. We evaluate the performance of diverse models on a diverse dataset (BCCC-cPacket-Cloud-DDoS-2024) with diverse samples, attack types, and network traffic patterns, including Artificial Neural Network (ANN), Long Short-Term Memory (LSTM), XGBoost, and LightGBM. Model evaluation is performed over several metrics like accuracy, precision, recall, F1 score, ROC-AUC. The findings suggest that ensemble learning models, which includes XGBoost and LightGBM, provide better accuracy and robustness than other models, which makes them suitable for the real-time detection tasks. Deep learning algorithms perform well when dealing with complex patterns in the data, but need more computational power. It also offers fundamental insights to help in selecting optimal IDS model(s) with respect to performance and resource constraints for real-time and offline detection.