End-to-end deep learning-based asymmetric encryption scheme for optical fiber communication

被引：0

作者：

Yao, Yueyi ^{[1
,2
,3
]}

Ma, Jie ^{[1
,2
,3
]}

Lu, Jia ^{[1
,2
,3
]}

Liu, Jianfei ^{[1
,2
,3
]}

Zeng, Xiangye ^{[1
,2
,3
]}

Luo, Mingming ^{[1
,2
,3
]}

机构：

[1] School of Electronics and Information Engineering, Hebei University of Technology, Tianjin,300401, China

[2] Tianjin Key Laboratory of Electronic Materials & Devices, Tianjin,300401, China

[3] Hebei Provincial Key Laboratory of Advanced Laser Technology and Equipment, Tianjin,300401, China

来源：

Optics Communications | 2025年 / 591卷

基金：

中国国家自然科学基金;

关键词：

Convolution - Cryptography - Fiber optic networks - Light transmission - Network security - Optical fibers - Signal processing;

D O I：

10.1016/j.optcom.2025.132203

中图分类号：

学科分类号：

摘要：

In this paper, a high-security asymmetric encryption based on end-to-end deep learning (AE-E2EDL) is innovatively proposed in optical fiber communication systems. AE-E2EDL leverages a deep convolutional generative adversarial network (DCGAN) for dynamic key generation and an autoencoder for asymmetric encryption, thereby eliminating the need for pre-shared keys, which are required in traditional symmetric encryption (SE) schemes. Compared to SE, the proposed AE-E2EDL significantly mitigates key distribution complexities while enhancing resistance to key exposure risks. Moreover, by integrating a signal-driven key generation strategy and pre-training optimization mechanism, the proposed scheme demonstrates better eavesdropping resistance, enhanced bit error rate (BER) performance, and faster convergence speed. Results show that the training time of the pre-trained system is 55 % shorter than that of the traditional system. In addition, the signal-based E2EDL key generation outperforms noise-based key generation by three orders of magnitude in secrecy capacity across 80-km fiber channels, thereby advancing secure data transmission in optical networks. © 2025

引用