Physical Layer Key Distribution Technology on Deep Learning

This study presents a deep learning-based physical layer key distribution scheme. The primary issues this scheme aims to address include the fluctuation of communication quality during continuous wireless communication, leading to unstable channel characteristics, and the inefficiency of physical la...

Full description

Saved in:
Bibliographic Details
Published in2023 International Conference on Networks, Communications and Intelligent Computing (NCIC) pp. 123 - 127
Main Authors Shi, Yuhao, Tang, Jie, Wang, Ruifei, Wen, Hong, Han-Ho, PIN, Chang, Shih Yu
Format Conference Proceeding
LanguageEnglish
Published IEEE 17.11.2023
Subjects
Costs
Deep learning
Fluctuations
Key Discrepancy Rate
Long Short-Term Memory Network
Memory management
Noise
Physical Layer Key
Quantization (signal)
Wireless communication
Online AccessGet full text

Cover

More Information
Summary:This study presents a deep learning-based physical layer key distribution scheme. The primary issues this scheme aims to address include the fluctuation of communication quality during continuous wireless communication, leading to unstable channel characteristics, and the inefficiency of physical layer key generation. It cleverly predicts channel characteristics through the introduction of key techniques like LSTM. Additionally, it increases the system's idle time for computation and communication, reducing the computational and communication costs during busy periods. The study utilizes previously collected channel characteristic data, enabling the system to more effectively acquire channel gains and generate a significant number of secure keys. Compared to traditional methods, this approach overcomes the limitation of generating only one key per channel probing and to some extent mitigates the issue of channel instability, significantly enhancing key distribution efficiency. Experimental validation demonstrates that this technology can predict channel characteristics to a considerable extent. Three quantization methods are employed for both original and predicted channel gains to derive physical layer keys. The conclusion from the verification and comparison is that the consistency between the predicted keys and the original keys exceeds 99.4%. This approach effectively addresses issues of poor key distribution stability and low efficiency.
DOI:10.1109/NCIC61838.2023.00027