Logo image
Artificial intelligence -based intrusion detection and prevention in IOT-enabled software-defined wireless sensor network environments
Dissertation   Open access

Artificial intelligence -based intrusion detection and prevention in IOT-enabled software-defined wireless sensor network environments

Joseph Byamungu Kipongo
Doctor of Philosophy (PHD), University of Johannesburg
Handle:
https://hdl.handle.net/10210/519332

Abstract

Intrusion detection systems (Computer security) Artificial intelligence Wireless sensor networks Internet of Things
Wireless sensor networks (WSNs) and the advancement into software-defined wireless sensor networks (SDWSNs) are crucial components in the rapid evolution of the Internet of Things (IoT), with applications demanding secure and efficient task execution. Despite their potential, these networks face significant security challenges due to their vulnerability to various attacks, making the implementation of effective intrusion detection systems (IDSs) and intrusion prevention systems (IPSs) essential. To address the challenges in IoT-enabled SDWSN environments, this research introduces a new model with a honeycomb structure-based intrusion detection and prevention system (IDPS). In this system, there is a hybrid IDS, secure authentication using the three-dimensional cube (3D cube) algorithm, modified honeycomb-based network partitioning, clustering, reinforcement learning (RL)-based intelligent routing with a transfer learning-based deep Q network (TLDQN), and honeycomb-based network partitioning. A driver training-based optimization (DTO) algorithm and a bidirectional generative adversarial network (Bi-GAN) algorithm are used as the last methods to detect malicious nodes and intrusions. Additionally, this research also focuses on preprocessing, feature extraction, and intrusion recognition using the network security laboratory-knowledge discovery in databases (NSL-KDD) dataset in order to evaluate intrusion detection in IoT-enabled SDWSNs. Feature extraction is facilitated by the novel exponential grey wolf-optimized grid search algorithm (EGWOGSA), while feature selection is achieved using the novel symmetric gradient Boruta for enhanced feature selection algorithm (SGBFSA). Intrusion detection is executed through a gated bidirectional recurrent convolutional neural network (GBR-CNN) algorithm. This research contributes to the advancement of security in IoT-enabled SDWSNs through advanced feature optimization and artificial intelligence (AI) techniques. The simulation results demonstrate the effectiveness of the proposed work in detecting network intrusions with high accuracy, outperforming existing methods across most measures, such as energy consumption, throughput, and computational overhead, while maintaining competitive performance in packet delivery ratio and network time, thereby demonstrating its effectiveness in securing IoT-enabled SDWSN environments.
pdf
KIPONGO JB – 220160236 (FINAL)11.96 MBDownloadView
Open Access

Metrics

1 File views/ downloads
3 Record Views

Details

Logo image