Cellular Neural Network Encryption Scheme for Time Synchronization and CPAs Resistance in OFDM-PON

Abstract

In this paper, we first experimentally demonstrate a novel cellular neural network (CNN)-based physical layer encryption scheme to achieve the capability of chosen-plaintext attacks (CPAs) resistance and time synchronization simultaneously in orthogonal frequency division multiplexing passive optical network (OFDM-PON). By utilizing the hyperchaotic phenomena within a certain parameter range characteristics of CNN, a four-dimensional (4-D) CNN-based system is constructed to strengthen the security of data transmission. And, according to CNN, the chaotic Feistel transform is executed after extracting the indexes of QAM data, in which completely dynamic encrypted data can be achieved for the CPAs resistance. Moreover, a chaotic training sequence generated by CNN used as timing sequence is added to the transmitted OFDM signals to obtain time synchronization and further increasing the confidentiality of the encryption system. A verified experimental system with 10-Gb/s 16-QAM encrypted OFDM signals through 20-km single-mode fiber (SMF) transmission is conducted. And, the results show that the Feistel and CNN-based physical-layer encryption scheme can generate completely dynamic ciphertexts for the CPAs resistance and accurately realize the time synchronization, and without apparent receiver sensitivity deterioration (0.3dB) is introduced in comparison with other against CPAs schemes.