Scholarlink

As the number of Internet of Things (IoT) devices has been exponentially increasing, the importance of network security has been established due to the growth in volume and increasing complexity of cyber-attacks. More conventional intrusion detection systems (IDS) are prone to fail to identify sophisticated or novel attacks, particularly within constrained resources’ IoT worlds. The aim of this project is to build an improved anomaly detection system for IoT networks using the Machine Learning (ML) and Deep-Learning (DL) models. In this solution, two benchmark datasets will be utilized as follows; NF-ToN-IoT has realistic IoT network traffic while UNSW-NB15 dataset is an intrusion detection dataset. The machine learning models applied include Random Forest, Decision Tree, Logistic Regression, Naive Bayes, XGBoost, and AdaBoost, along with a stacking ensemble. These were compared against deep learning approaches such as Autoencoders, CNNs, and LSTM networks. A performance analysis was conducted on all models according to the metrical standards of accuracy, precision, recall, F1 score and ROC AUC. Further studies are required to appreciate the potential of a combined ML-DL hybrid model architecture and optimization strategies for real-time implementation.

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

Enhanced Anomaly Detection in IoT Networks Using Machine Learning and Deep Learning Approaches

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Nishthaa Jain,
  • Surbhi Sharma,
  • Renu Dalal,
  • Manju Khari,
  • Arvind Panwar

摘要

As the number of Internet of Things (IoT) devices has been exponentially increasing, the importance of network security has been established due to the growth in volume and increasing complexity of cyber-attacks. More conventional intrusion detection systems (IDS) are prone to fail to identify sophisticated or novel attacks, particularly within constrained resources’ IoT worlds. The aim of this project is to build an improved anomaly detection system for IoT networks using the Machine Learning (ML) and Deep-Learning (DL) models. In this solution, two benchmark datasets will be utilized as follows; NF-ToN-IoT has realistic IoT network traffic while UNSW-NB15 dataset is an intrusion detection dataset. The machine learning models applied include Random Forest, Decision Tree, Logistic Regression, Naive Bayes, XGBoost, and AdaBoost, along with a stacking ensemble. These were compared against deep learning approaches such as Autoencoders, CNNs, and LSTM networks. A performance analysis was conducted on all models according to the metrical standards of accuracy, precision, recall, F1 score and ROC AUC. Further studies are required to appreciate the potential of a combined ML-DL hybrid model architecture and optimization strategies for real-time implementation.