Intrusion Detection System for IoT based on Complex Networks and Machine Learning

Network Intrusion Detection Systems (NIDS) are critical tools for detecting and preventing cyber-attacks in computer networks. Traditional rule-based NIDS can be limited in their ability to detect complex and sophisticated attacks. However, recent advancements in Artificial Intelligence (AI), such as deep learning algorithms, have provided new methods for enhancing NIDS capabilities, improving detection rates, reducing false positives, and adapting them to new and emerging threats in real-time. However, the majority of the existing methods in the literature do not take into consideration information about the topological properties of the complex network that could model any network traffic data. Complex networks refer to networks or graphs in which the connections between the nodes or entities are not randomly distributed, but instead exhibit some kind of structure or pattern (in our case, nodes can represent the IoT devices while the connections represent the data exchanged between them). Complex network features, which refer to advanced analytical techniques that can identify hidden patterns and relationships in network traffic data, have emerged as a powerful tool for enhancing NIDS capabilities. By using these features, NIDS can detect and adapt to new types of attacks, improving accuracy and effectiveness. This paper presents an overview of NIDS using AI and complex network features, that can detect and adapt to new types of attacks. We discuss the importance of using complex network features for NIDS and provide an example of how these features can enhance the accuracy and effectiveness of NIDS. Our results have shown that our method outperforms other machine learning techniques in terms of accuracy, and indicate its potential to improve intrusion detection. Overall, this paper highlights the importance of incorporating AI and complex network features into NIDS to improve the detection and prevention of cyber-attacks in computer networks.