RESEARCH ON IIOT SECURITY: NOVEL MACHINE LEARNING-BASED INTRUSION DETECTION USING TCP/IP PACKETS

Abstract

The Industrial Internet of Things (IIoT) explosive expansion has raised questions regarding the safety of industrial systems. Networks like these are crucially protected from a variety of cyber threats by intrusion detection systems (IDSs). In order to detect intrusions in the IIoT environment utilizing TCP/IP packets, this work introduces a novel Hybrid Deep Convolutional Autoencoder and Splinted Decision Tree (HDCA-SDT) technique. High-level features are extracted from the unprocessed TCP/IP packet data using the DCA. The retrieved features are then classified using the SDT algorithm into various intrusion categories. In order to enable quicker decision-making yet preserve accurate results, the SDT technique effectively divides the feature space. The NSL-KDD dataset is used to train and assess the model. The efficiency of the suggested hybrid strategy is shown by experimental findings. Comparing the proposed hybrid approach to conventional intrusion detection methods, it acquired higher detection accuracy. The model also demonstrates robustness to fluctuations in traffic on the network and possesses the ability to identify known and unidentified intrusions with high recall rates.

Keywords

• internet of things (iot)
• industrial systems
• intrusion detection system (ids)
• tcp/ip packet
• security
• hybrid deep convolutional autoencoder
• and splinted decision tree (hdca-sdt)