Enhanced Security Authentication Based on Convolutional-LSTM Networks.

Area of Science:

• Cybersecurity
• Machine Learning
• Wireless Communications

Background:

• Classical security authentication models struggle with imperfect channel estimation and time-varying links.
• Statistical decision methods for physical layer authentication face challenges in dynamic, non-stationary environments.

Purpose of the Study:

• To develop a deep learning-based authentication approach for physical layer security.
• To improve the adaptability and convergence of authentication in dynamic communication environments.

Main Methods:

• Introduced a deep learning-based authentication approach.
• Designed an intelligent detection framework using a Convolutional-Long Short-Term Memory (Convolutional-LSTM) network.
• Analyzed the robustness and detection performance of the learning authentication scheme.

Main Results:

• The Convolutional-LSTM network effectively handles channel differences without prior statistical knowledge.
• The proposed scheme demonstrates improved adaptability and convergence in physical layer authentication.
• Extensive simulations and experiments show significantly improved detection accuracy in time-varying environments.

Conclusions:

• Deep learning-based authentication offers a robust solution for physical layer security in dynamic environments.
• The Convolutional-LSTM network is effective for learning and tracking channel variations.
• The proposed approach enhances the performance and reliability of wireless security authentication.