A Malicious Domain Detection Model Based on Improved Deep Learning.

Area of Science:

• Cybersecurity
• Machine Learning
• Network Security

Background:

• Malicious domain names present significant threats to network and social security.
• Existing research on malicious domain detection has yielded various results.
• The internet's rapid expansion amplifies the risks associated with malicious domains.

Purpose of the Study:

• To propose an improved deep learning model for enhanced malicious domain name detection.
• To leverage the strengths of multiple neural network architectures for superior detection performance.
• To outperform existing single or dual-model approaches in identifying malicious domains.

Main Methods:

• Development of an improved deep learning model integrating Convolutional Neural Network (CNN), Temporal Convolutional Network (TCN), and Long Short-Term Memory (LSTM).
• Comparative analysis of the proposed model against combined CNN-LSTM and CNN-TCN models.
• Experimental evaluation of detection accuracy and regression rates.

Main Results:

• The proposed integrated deep learning model demonstrated superior performance compared to combined CNN-LSTM and CNN-TCN models.
• Achieved a high accuracy rate of 99.76% in malicious domain detection.
• Attained a regression rate of 98.81%.

Conclusions:

• The improved deep learning model effectively enhances malicious domain name detection capabilities.
• Combining CNN, TCN, and LSTM offers significant advantages over traditional or dual-model methods.
• The model provides a robust solution for improving network security against malicious domains.