High-security three-dimensional CAP modulation technique integrating end-to-end constellation optimization and four-level masking

2025-03-01

PRE

Juntao Zhang

刘

刘波 (Bo Liu) *

Jianxin Ren

Yaya Mao

Xiangyu Wu

Rahat Ullah

Jie Cui

Yilan Ma

Jianye Zhao

Dongdong Xu

Shuaidong Chen

Shuyu Zhou

DOI：10.3788/COL202523.030601

Original

Original request for help

Save

Abstract

En 中文

In this paper, a high-security three-dimensional carrierless amplitude and phase (3D-CAP) modulation technique is proposed, integrating deep learning with four-level masking. The 3D constellation geometry is optimized using an autoencoder (AE) with an additive white Gaussian noise (AWGN) channel model, reducing complexity by 40% compared to a variational autoencoder (VAE). Experimental validation on a 2 km seven-core fiber intensity modulation/direct detection (IM/DD) system shows a 1 dB improvement in receiver sensitivity. A 3D chaotic oscillator model is used for chaotic selective mapping, polynomial-like masking, constellation rotation, and subcarrier masking. The encrypted 3D-CAP signal achieves a key space of up to 10103, with strong anti-noise and confidentiality performance.

Keywords:

autoencoder

geometric shaping

high security

chaotic encryption

High-security three-dimensional CAP modulation technique integrating end-to-end constellation optimization and four-level masking

Abstract

Journal

Organization

Cited Papers

Citing Papers