A Spatiotemporal-Oriented Deep Ensemble Learning Model to Defend Link Flooding Attacks in IoT Network.

Area of Science:

• Cybersecurity
• Network Security
• Artificial Intelligence

Background:

• Link Flooding Attacks (LFAs) are a type of distributed denial-of-service (DDoS) attack targeting backbone links.
• Traditional defenses against LFAs are often heuristic and fail to adapt to evolving attack patterns.
• Existing AI methods detect LFAs but lack spatiotemporal pattern analysis and defense recommendations.

Purpose of the Study:

• To develop an advanced deep ensemble learning model for robust defense against LFAs.
• To address the limitations of traditional and existing AI-based LFA detection methods.
• To integrate LFA detection with real-time mitigation strategies.

Main Methods:

• Designed a Stacking-based integrated Convolutional Neural Network-Long Short-Term Memory (SCL) model.
• Utilized continuous network status as input for CNN to capture spatiotemporal attack features.
• Employed LSTM for dynamic pattern review and obsolete pattern elimination, enhancing decision accuracy.
• Integrated System Detector and LFA Mitigator modules for simultaneous detection and mediation.

Main Results:

• The SCL model demonstrated a 92.95% accuracy rate in successfully blocking LFAs.
• Achieved a 60.81% improvement in accuracy compared to traditional defense methods.
• Validated the effectiveness of the deep ensemble approach in handling complex network attacks.

Conclusions:

• Deep ensemble learning shows significant potential for enhancing network security against sophisticated attacks like LFAs.
• The SCL model offers a promising direction for developing adaptive and intelligent network defense systems.
• Further research into deep ensemble learning is recommended for advancing cybersecurity solutions.