VMFCVD: An Optimized Framework to Combat Volumetric DDoS Attacks using Machine Learning.

Area of Science:

• Computer Science
• Cybersecurity
• Network Security

Background:

• Distributed Denial of Service (DDoS) attacks cause significant downtime, reputational damage, and revenue loss.
• Volumetric DDoS attacks, often amplified by IoT devices, pose a major threat to network availability.
• Existing DDoS defense systems require enhancement to effectively mitigate large-scale attacks.

Purpose of the Study:

• To propose a novel voting-based multimode framework to combat volumetric DDoS attacks (VMFCVD).
• To enhance the speed and accuracy of DDoS attack detection and mitigation.
• To evaluate the effectiveness of the proposed framework against various DDoS and botnet datasets.

Main Methods:

• Development of VMFCVD, a framework utilizing three modes: Fast Detection Mode (FDM), Defensive Fast Detection Mode (DFDM), and High Accuracy Mode (HAM).
• FDM employs highly dimensionally reduced datasets for accelerated traffic classification during attacks (over 97% reduction, 99.9% accuracy).
• DFDM enhances detection accuracy by refining FDM's techniques, while HAM provides superior accuracy during stable server conditions.

Main Results:

• VMFCVD demonstrated superior performance compared to recent studies on multiple benchmark datasets (CICIDS2017, CSE-CIC-IDS2018, CICDDoS2019, etc.).
• FDM achieved significant dimension reduction while maintaining high detection accuracy.
• The framework performed exceptionally well under simulated DDoS attack conditions.

Conclusions:

• VMFCVD offers an effective solution for mitigating volumetric DDoS attacks.
• The multimode approach balances detection speed and accuracy, crucial for real-time network defense.
• The proposed framework represents a significant advancement in combating large-scale DDoS threats.