Deep Q-Learning-Based Buffer-Aided Relay Selection for Reliable and Secure Communications in Two-Hop Wireless Relay...

Area of Science:

• Wireless Communications
• Network Security
• Machine Learning Applications

Background:

• Wireless signals are prone to fading and eavesdropping, compromising communication integrity.
• Existing relay selection schemes often address either reliability or security, but not both.
• Buffer-aided amplify-and-forward (AF) networks require robust strategies for dependable and confidential data transmission.

Purpose of the Study:

• To propose a novel buffer-aided relay selection scheme integrating both reliability and security.
• To address the limitations of existing schemes that focus on single aspects of communication performance.
• To enhance the overall performance of two-hop AF wireless networks against fading and eavesdropping.

Main Methods:

• Development of a buffer-aided relay selection scheme utilizing deep Q-learning (DQL).
• Monte Carlo simulations to evaluate the proposed scheme's performance.
• Comparison with benchmark schemes, including the max-ratio scheme.

Main Results:

• The proposed DQL-based scheme effectively achieves reliable and secure communications in two-hop AF networks.
• The scheme demonstrates superior performance in terms of secrecy outage probability (SOP) compared to the max-ratio scheme.
• Monte Carlo simulations validate the scheme's effectiveness in mitigating connection outage probability (COP) and SOP.

Conclusions:

• The proposed deep Q-learning-based buffer-aided relay selection scheme offers a comprehensive solution for reliable and secure wireless communications.
• This approach successfully balances the competing demands of communication integrity and data confidentiality in AF networks.
• The findings highlight the potential of DQL in optimizing complex wireless network functionalities.