A DoS attack detection method based on adversarial neural network.

Area of Science:

• Cybersecurity
• Network Security
• Artificial Intelligence

Background:

• Denial-of-Service (DoS) attacks pose significant threats to network availability.
• Existing DoS detection methods have limitations in accuracy and speed.
• Deep learning offers potential for advanced network traffic analysis.

Purpose of the Study:

• To analyze the influence of deep learning on DoS attack detection.
• To propose a novel deep learning model for enhanced DoS attack identification.
• To evaluate the proposed model's performance on a standard intrusion detection dataset.

Main Methods:

• Examined DoS attack concepts, strategies, and current detection methodologies.
• Developed a deep learning-based distributed DoS attack detection system.
• Proposed the Improved Conditional Wasserstein Generative Adversarial Network with Inverter (ICWGANInverter) model.
• Utilized reconstruction error for classification and tested on the NSL-KDD dataset.

Main Results:

• The ICWGANInverter model demonstrated excellent detection performance with Area Under the ROC Curve (AUC) values above 0.8.
• Mean square error of continuous feature reconstruction increased with noise factor in tested sub-datasets.
• Achieved a detection accuracy of 87.79%, surpassing other models.

Conclusions:

• The ICWGANInverter model offers superior performance for detecting DoS attacks.
• The proposed deep learning approach provides significant benefits for network security.
• The model effectively identifies incomplete network traffic patterns characteristic of DoS attacks.