In this paper, 10 Gb/s free space optical (FSO) wiretap channel using optical code division multiple access (OCDMA) time-diversity reception is designed and demonstrated for the first time. Reliability and security are investigated under weak and strong turbulence conditions. Two-dimensional optical encoder and decoder are constructed by wavelength selective switches and optical delay lines, and one-dimensional optical encoder and decoder are constructed by couplers and tunable optical delay lines. Experimental results show that reliability and security can be enhanced simultaneously using OCDMA time-diversity reception. At the received power of 2.58dBm in strong turbulence, compared to non-diversity system, bit error rate (BER) of legitimate user in OCDMA time-diversity system is reduced from $3.9\times 10^{-7} to $6.73 \times 10^{-8}$ . Furthermore, the reliability improvement is more obvious when the correlation between the two signals of OCDMA time-diversity is smaller. The secrecy capacity can be increased from 0.546 bit/symbol to 0.665 bit/symbol in strong turbulence when received power is 2.28dBm. On the other hand, the secrecy capacity of strong turbulence is about 15% higher than that of weak turbulence.

Due to the effect of atmospheric turbulence, the transmission reliability and physical-layer security of free space optical (FSO) communication system are facing many challenges. In this paper, for the first time, we design and investigate a 10 Gb/s reconfigurable FSO wiretap channel based on spatial diversity and optical code division multiple access (OCDMA). Reconfigurable two-dimensional optical encoder and decoder are constructed by wavelength selective switch and optical delay lines, and reconfigurable one-dimensional optical encoder and decoder are constructed by couplers and tunable optical delay lines. Under three different turbulence conditions, bit error rates (BER) of legitimate user and eavesdropper are measured. It can be seen that under turbulence effects, BER of legitimate user in spatial-diversity is lower than that of non-diversity system. By the rapid code reconfiguration of legitimate users, the eavesdropper can only use unmatched decoder. Thus, the eavesdropping methods of code interception and brute-force searching can be avoided. Experimental results show that this scheme can ensure the reliable transmission of legitimate users, and the eavesdropper can't crack the information when the extraction ratio is 1%. Hence, reliability and physical-layer security can be enhanced simultaneously.

A multi-user cross-layer security network based on optical code division multiple access (OCDMA) and algorithmic cryptography is proposed, and its security performance is analyzed quantitatively for the first time. In this scheme, eavesdroppers have to not only intercept the optical code of OCDMA in physical-layer, but also break the algorithmic cryptography in data-layer. Thus, the information security will be enhanced effectively. Synchronous transmission system and asynchronous transmission system are analyzed respectively. It can be seen that the cross-layer security of asynchronous transmission is far greater than that of synchronous transmission when the number of users is large enough. OptiSystem simulation is used to investigate the image transmission in cross-layer security system. The transmission distance of single mode fiber is 100 km, and transmission rates are 10 Gbit/s and 1 Gbit/s respectively. It can be seen that legitimate user can recover original image correctly with matched decoder and correct encryption algorithm. However, eavesdroppers cannot recover the image correctly with matched decoder/wrong encryption algorithm or unmatched decoder/correct encryption algorithm. Therefore, our scheme can greatly enhance the security of image transmission system.

The automatic compliance review of substation design model greatly depends on the computer-understandable layout criteria. Unfortunately, most of these layout criteria are depicted in a series of design codes written in natural language. Meanwhile manual conversion from unstructured code text to layout rules is often costly and time consuming. In this regard, automatic extracting layout information can be developed utilizing Natural Language Processing (NLP) tools based on deep learning. The layout texts are collected and manually labeled to establish the experimental corpus for further study, and then the NLP-assisted information extraction method is developed mainly using the Bidirectional Long Short-Term Memory (BiLSTM) networks and Conditional Random Field (CRF). Finally, the information extraction model is trained on the labeled experimental corpus. The test results indicate that the developed model is feasible to extract semantic information of layout constraints residing in design code.