Multi-Layered Security Framework for Intrusion Detection System in Software Defined Networking Environment Using Machine Learning

Abstract

In Software-Defined Networking (SDN), Intrusion Detection Systems (IDSs) are crucial for enhancing network security. These systems analyze and detect network anomalies dynamically, making SDN environments more responsive to emerging threats. The BAT-MC model has demonstrated effectiveness in improving network traffic analysis within SDN. This model integrates bidirectional context understanding and attention mechanisms, combining Bi-directional Long ShortTerm Memory (BiLSTM) and Attention layer with multiple convolutional layers. The proposed system offers precise anomaly detection and improved traffic optimization within the SDN infrastructure. Traditional IDSs relying on the NSL-KDD dataset face accuracy and scalability constraints, even with machine learning techniques. Moreover, manual feature engineering complicates the challenge due to the increasing diversity and complexity of network traffic. To tackle these issues, BAT-MC employs advanced deep learning techniques by seamlessly integrating BiLSTM and attention mechanisms, eliminating the need for manual feature design and significantly improving intrusion detection capabilities. To address existing dataset limitations, the In-SDN dataset is used, aiming to enhance system performance in intrusion detection when combined with the BAT-MC model and ensemble methods. Through a comprehensive approach, the goal is to improve accuracy, scalability, and adaptability, achieving an impressive $\mathbf{8 6 \%}$ performance in safeguarding SDN environments from various potential threats.