基于机器学习储备池计算的混沌保密通信机制设计与实现

Design and realization of chaotic secure communication based on machine learning approach⁃reservoir computing

针对传统混沌保密通信中安全性低、耦合机制复杂、接收端须配备特定动力学系统等问题,提出了一种基于机器学习——储备池计算(Reservoir Computing)的混沌保密通信设计,并通过数值与实验方法验证了机制的创新性与可行性。在数值方面,以图像和语音传输为例,采用具有更高安全性的时滞Lorenz系统和Mackey⁃Glass模型进行加密,在加密系统信息完全未知的前提下,接收端经训练后,可实现与发送端混沌完全同步并进行信息解密。在实验方面,基于可编程逻辑芯片(FPGA)对机制进行了硬件设计与实现,并以实时视频传输为例,验证了该保密通信设计的实际应用价值。还引入了针对不同加密系统的校准方法,以进行长时间稳定混沌同步和通信。此外,通过噪声研究,证明了该通信机制的鲁棒性。理论与实验结果对推动新型保密通信与机器学习交叉研究有一定的参考价值。

Since the receiver in traditional chaos⁃based secure communication often faces the problems such as low security,complicated coupling and design constrains,a novel scheme using machine learning approach⁃reservoir computing is proposed with both numerical and experimental investigation conducted.Numerically,the secure communication of image and voices are studied by employing the time⁃delayed Lorenz and Mackey⁃Glass as encrypting systems.Thus,the receiver can decrypt the information successfully without any pre⁃knowledge of the encrypted system.Experimentally,the digital implementation using the field programmable gate array(FPGA)is conducted and tested with real⁃world video transmission.Different calibration methods are introduced for achieving stable and long⁃term communication.In addition,the noise effect is considered to demonstrate the noise⁃robustness of the proposed method.The result is of great significance for pushing the development of smart signal processing techniques.