全光混沌由于产生结构简单和混沌复杂度高等特点，在混沌保密通信中起着重要的作用。由于全光混沌需要发射端与接收端激光器参数几乎完全一致才可达到很好的同步效果，一直没有在保密通信方向取得很好的应用。文章提出了一种利用神经网络对激光器进行建模同步的方法，在发射端利用大量已知混沌信号和混沌加密信号对神经网络进行训练，得到一个与发射端激光器结构相似的神经网络模型，在接收端利用此神经网络模型对加密信息进行解密，并通过实验成功对5 Gbit/s信号进行了混沌加密和解密。此方法大大降低了全光混沌同步的复杂度，并为点到多点全光混沌通信提供了可能。

All-optical chaotic communication plays an important role in chaotic communication due to its simple structure and high complexity. Chaos synchronization is the most important part in chaotic optical communication, requiring the perfect parameter matching of the chaotic transmitter and receiver. Due to the difficulty of perfect devices matching, the practical applications of chaotic optical communications are limited. This paper proposed a method of modeling the chaotic laser using neural network, which can significantly simplify the structure of the chaotic receiver while keeping the system security. A 5 Gbit/s signal transmission of non-return-to-zero message hidden in a chaotic optical carrier is successfully experimentally demonstrated. This method greatly reduces the complexity of all-optical chaos synchronization and provides the possibility of point-to-multipoint all-optical chaotic communication.