Abstract:
"The Internet of Things (IoT) has improved sectors like healthcare and smart cities but also
made them vulnerable to DDoS attacks. Traditional IDS are ineffective in IoT networks due to
their diversity and resource-constraint nature. This research proposes a solution using a Multi-Agent System and ensemble machine learning techniques, specifically focusing on Naive Bayes classifiers, to detect and prevent DDoS attacks effectively.
This study introduces an IDS capable of efficiently identifying anomalies indicative of DDoS
attacks within IoT infrastructures by integrating a Naive Bayes ensemble technique with MAS.
The ensemble method enhances detection accuracy through collective intelligence, while the distributed nature of MAS ensures scalability and efficient resource utilisation. The research adopts a pragmatic methodology, combining literature review, expert interviews, and prototyping, to refine the IDS's capabilities iteratively. This allows for the practical evaluation of the IDS in simulated IoT environments and contributes to the iterative improvement of the detection model through real-world feedback loops.
The ensemble Naïve Bayes model achieved 77% accuracy. It efficiently handles large datasets and is adaptable to IoT environments, making it a viable solution for real-time intrusion detection. MAS agents have a single responsibility and can be distributed within IoT networks, consuming fewer resources. Advanced features like agent mobility can be utilised when more resources are needed."
