Customised Intrusion Detection for an Industrial IoT Heterogeneous Network Based on Machine Learning Algorithms...

Area of Science:

• Computer Science
• Network Security
• Internet of Things

Background:

• The Internet of Things (IoT) relies on the Routing Protocol for Low-Power and Lossy Networks (RPL) for connectivity.
• RPL security is critical due to increasing IoT reliance, but existing intrusion detection systems (IDS) struggle with heterogeneous IoT environments and novel threats.
• Unified IDS models are insufficient for detecting new RPL intrusions in diverse IoT networks.

Purpose of the Study:

• To propose a novel federated transfer learning-assisted customized distributed IDS (FT-CID) model for detecting RPL intrusions in heterogeneous IoT environments.
• To enhance the security of the RPL protocol by enabling the detection of both existing and novel intrusions.
• To address the limitations of existing unified IDS solutions in complex IoT ecosystems.

Main Methods:

• Dataset collection and simulation of normal and abnormal traffic in Contiki-NG OS for RPL-IIoT.
• Federated transfer learning (FTL) for training edge IDSs using local and globally shared parameters.
• Customized IDS model development through partial retraining and leveraging shared server knowledge.

Main Results:

• The FT-CID model successfully detects RPL intrusions in heterogeneous IoT networks.
• The proposed model achieved an accuracy of 85.52% in detecting intrusions.
• Implicit utilization of local and global parameters from diverse IoTs enhanced overall RPL security.

Conclusions:

• The FT-CID model offers a customized and effective solution for RPL intrusion detection in heterogeneous IoT.
• Federated transfer learning is a viable approach for building adaptive and secure IoT IDS.
• The FT-CID model significantly improves RPL security by addressing the challenge of novel intrusion detection.