Block chain assisted deep learning framework for intrusion detection in Zigbee based IoT network

The adoption of Internet of Things (IoT) technologies in smart environments is growing, and this requires reliance on energy-saving communication protocols, including Zigbee. This ubiquity naturally increases the cyber-attack area, which requires innovative security measures countering the use of resource-limiting devices. The current study presents a new real-time Intrusion Detection System (IDS) model based on a strategic implementation of both deep learning (DL) to detect threat and blockchain technology to provide immutable audit log in a Zigbee-based smart home. Our approach involves the use of a carefully annotated data set based on the analysis of the IEEE 802.15.4 at the packet level through analysis of Packet Capture (PCAP) files of human motion events. Our comparative analysis of five deep learning architectures, including Convolutional Neural Networks (CNN), Bidirectional Long Short-Term Memory (BiLSTM), Gated Recurrent Units (GRU), Simple Recurrent Neural Network (RNN), and GRU+LSTM hybrid is very rigorous, and this paper provides the best model to detect the presence of an anomalous network behavior pattern. The findings justify the effectiveness of the framework because it is evident that the proposed IDS was highly effective with an accuracy of 99.77 per cent and an ROC-AUC of 0.9977 as applied to the argeted Zigbee data. With this performance, a strong ability to identify well-known patterns of attacks is formed. Implementation of a lightweight private blockchain will make the records of the detention safe, publicly-readable, and non-deniable. GRU+LSTM always showed better overall knowledge in terms of detection and thus it has been the best architecture to be deployed. These results indicate that the challenges of deep learning and blockchain provide an attractive and computationally efficient paradigm of improving security and forensic function in the resource-limited IoT. The essence of the originality of this work is in the fact that it provides quantitative metrics of performances of a lightweight Proof-of-Authority (PoA) blockchain, coupled with an optimized and tailored hybrid DL model, which is specifically verified on Zigbee PCAP data of micro-scale and frame-by-frame analysis.