Enhancing AI-Driven Intrusion Detection in SDN: Exploring the Power of Dimensionality Reduction Techniques

The increasing security of Software-Defined Networking (SDN) environments underlines the compelling need for more efficient and adaptive Intrusion Detection Systems (IDS). In this paper, we enhance AI-driven intrusion detection in SDN by exploiting advanced dimensionality reduction methods such as Principal Component Analysis (PCA), Uniform Manifold Approximation and Projection (UMAP), t-distributed Stochastic Neighbor Embedding (t-SNE) and Linear Discriminant Analysis (LDA). We visualized network traffic data through scatter plots and subsequently analyzed the results to identify patterns and classes. We also generated reduced datasets for training various machine learning models, after which we conducted a thorough interpretation of performance metrics to evaluate the efficiency of the models. In the final phase of our research, we employed the Random Search method to optimize the UMAP parameters, thereby improving the performance of machine learning models that initially exhibited lower performance. Our research demonstrates the ability of dimensionality reduction techniques to refine the essential characteristics of the data, facilitating the accurate identification of malicious traffic and significantly improving IDS performance. This research sheds new light on the application of dimensional data analysis to enhance the security of SDN networks.