Machine-learning-based optical spectrum feature analysis for DoS attack detection in IP over optical networks.

Area of Science:

• Network Security
• Optical Communications
• Machine Learning

Background:

• Software-defined IP over optical networks are vulnerable to Denial of Service (DoS) attacks.
• Traditional security methods may not be sufficient for detecting sophisticated network threats.

Purpose of the Study:

• To develop and evaluate a novel machine learning-based approach for DoS attack detection in IP over optical networks.
• To analyze optical spectrum features for identifying network security anomalies.

Main Methods:

• Utilized machine learning algorithms (XGBoost, LightGBM, BP neural network) to analyze optical spectrum data.
• Collected datasets through numerical simulations and experimental trials.
• Assessed algorithm performance based on detection accuracy.

Main Results:

• All evaluated machine learning algorithms exceeded 97% detection accuracy for DoS attacks.
• The BP neural network achieved the highest accuracy: 99.55% in simulations and 99.74% in experiments.
• Optical spectrum data analysis proved effective for early DoS attack detection.

Conclusions:

• The proposed machine learning approach offers a promising new method for DoS attack detection in optical networks.
• Analyzing optical spectrum features enhances early warning capabilities for network security.
• The BP neural network demonstrates superior performance in this detection task.