Multi-Stage Learning Framework Using Convolutional Neural Network and Decision Tree-Based Classification for...

Area of Science:

• Computer Science
• Cybersecurity
• Industrial Control Systems

Background:

• Traditional Supervisory Control and Data Acquisition (SCADA) systems face challenges in flexibility, scalability, and management due to conventional network structures.
• Software-Defined Networking (SDN) offers potential solutions by separating control and data planes, but introduces new vulnerabilities, particularly against Distributed Denial of Service (DDoS) attacks targeting its centralized controller.
• Effective detection of DDoS attacks is crucial to prevent severe disruptions in SDN-based SCADA environments.

Purpose of the Study:

• To propose and evaluate a novel multi-stage learning model for the effective detection of DDoS attacks in SDN-based SCADA systems.
• To address the security vulnerabilities introduced by integrating SDN into SCADA networks.
• To enhance the resilience of critical industrial infrastructure against sophisticated cyber threats.

Main Methods:

• Development of a multi-stage learning model combining a 1-dimensional Convolutional Neural Network (1D-CNN) and decision tree-based classification.
• Creation of a new dataset featuring diverse attack scenarios within a specific experimental network topology for training and testing.
• Experimental validation of the proposed model's performance in detecting DDoS attacks.

Main Results:

• The proposed multi-stage learning model achieved a high accuracy rate of 97.8% in detecting DDoS attacks.
• The model demonstrated effective identification of various attack scenarios within the SDN-based SCADA network.
• Early detection capabilities were highlighted, enabling timely security measure implementation.

Conclusions:

• The developed multi-stage learning model is highly effective for detecting DDoS attacks in SDN-based SCADA systems.
• This approach offers a significant advancement in securing critical industrial control infrastructure.
• The findings underscore the potential of advanced machine learning techniques for robust cybersecurity in industrial environments.