ML-Based Detection of DDoS Attacks Using Evolutionary Algorithms Optimization.

Area of Science:

• Computer Science
• Cybersecurity
• Artificial Intelligence

Background:

• Modern reliance on information and communication technology makes systems vulnerable to cyber-attacks.
• Distributed Denial-of-Service (DDoS) attacks are a prevalent threat, causing significant downtime and financial losses.
• Existing DDoS detection methods show a decline in quantity and prior detection rates.

Purpose of the Study:

• To introduce an innovative approach for DDoS attack detection by integrating evolutionary optimization algorithms and machine learning.
• To propose and evaluate XGB-GA Optimization, RF-GA Optimization, and SVM-GA Optimization methods.
• To enhance the accuracy and robustness of DDoS attack detection systems.

Main Methods:

• Utilized datasets pertaining to DDoS attacks for training machine learning models (XGB, RF, SVM).
• Employed Evolutionary Algorithms (EAs) Optimization with Tree-based Pipelines Optimization Tool (TPOT)-Genetic Programming for model optimization.
• Implemented 10-fold cross-validation to assess model performance.

Main Results:

• Achieved high accuracy scores: 99.99% with XGB-GA, 99.50% with RF-GA, and 99.99% with SVM-GA.
• TPOT identified XGB-GA as the optimal algorithm for constructing the machine learning model.
• Demonstrated significant improvement over existing DDoS detection techniques.

Conclusions:

• The proposed XGB-GA, RF-GA, and SVM-GA methods offer a robust and accurate approach to DDoS attack detection.
• This research advances the field of cybersecurity by enhancing the resilience of digital infrastructures.
• The integration of evolutionary algorithms and machine learning provides a powerful tool for combating pervasive cyber threats.