Multi-wavelength optical secret sharing system and method based on optical diffraction neural network

Abstract

The invention relates to a multi-wavelength optical secret sharing system and method based on an optical diffraction neural network. The multi-wavelength optical secret sharing system comprises an input layer module, a modulation layer module and an output layer module which are connected in sequence. The input layer module is used for generating incident light beams with different wavelengths; the modulation layer module is used for carrying out phase modulation on an incident light beam and encrypting to-be-transmitted information split into a plurality of target images through optical diffraction calculation, the modulation layer module comprises a plurality of modulation layers, the on-off state of each modulation layer is independently controllable, and the modulation layer is used for carrying out phase modulation on the to-be-transmitted information. The trained modulation layer module can generate a unique corresponding secret key for each target image, and each secret key comprises the wavelength of an incident beam and the on-off state of each modulation layer; and the output layer module is used for reproducing the split information to be transmitted to realize optical secret sharing, the output layer module comprises a plurality of output planes, and each output plane is used for reproducing a corresponding target image according to different key combinations. Compared with the prior art, the security and capacity of information transmission can be further improved.

Description

Technical Field

[0001] The present invention belongs to the field of information security technology, and in particular relates to a multi-wavelength optical secret sharing system and method based on optical diffraction neural network. Background Art

[0002] Optical secret sharing technology is an information encryption technology that splits and stores secret information. Only when enough secret holders combine the information they hold can the final encrypted information be obtained. It is an important means of information security and data confidentiality and is widely used in personal information security, intelligent manufacturing, identity authentication and other fields.

[0003] The document "Multi-dimensional multiplexing optical secret sharing framework with cascaded liquid crystal holograms" (Keyao Li, Yiming Wang, Dapu Pi, Baoli Li, Haitao Luan, Xinyuan Fang, Peng Chen, Yanqing Lu and Min Gu, Opto-Electron. Adv. 7, 230121-1 (2024)) proposes an optical secret sharing system with cascaded liquid crystal holograms, which uses cascaded metasurfaces to split encrypted information and realize decentralized storage of information. Only when the number of metasurface holograms as keys is complete and correctly stacked, the encrypted information can be read in the corresponding polarization channel. The system uses the polarization state of the incident light, the external voltage and spacing applied to the cascaded liquid crystal shares as keys, and enhances the security of the information through the above-mentioned multidimensional multiplexing technology. However, the wavelength of the incident light beam used in the document is single, the number of modulation layers is small, and the generated image is distributed in a single plane. The capacity and security of the encrypted information need to be further improved. The document "Flat multifunctional liquid crystal elements through multi-dimensional information multiplexing" (Dongliang Tang, Zhenglong Shao, Xin Xie, Yingjie Zhou, Xiaohu Zhang, Fan Fan and Shuangchun Wen, Opto-Electron. Adv. 6, 220063-1 (2023)) proposes to use the polarization, wavelength and imaging distance parameters of the incident light as keys to achieve different display functions in the near field and far field. There are many types of keys used in this document, but the number of modulation layers used is a single layer, which does not solve the above problems well.

[0004] Optical diffraction neural network is a computing system that uses the diffraction properties of light to simulate the operation of neural networks, and has the advantages of high speed and low power consumption. The Chinese patent application with application publication number CN113822424A discloses an all-optical diffraction neural network system based on a metasurface, which consists of an incident object beam, a metasurface and a CMOS image processor. The CMOS image processor observes the light intensity of different incident object beams in the detection area to achieve classification and recognition functions. Since incident beams of different polarization states (wavelengths) carry different information, multifunctional classification and recognition functions can be achieved. Optical diffraction neural networks provide new ideas and technical support for improving existing optical secret sharing systems, but there is currently no direct use of optical diffraction neural networks for optical secret sharing systems. Therefore, it is urgent to design an optical secret sharing scheme based on optical diffraction neural networks to further improve the security and capacity of information transmission. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a multi-wavelength optical secret sharing system and method based on optical diffraction neural network, so as to further improve the security and capacity of information transmission.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] The present invention provides a multi-wavelength optical secret sharing system based on an optical diffraction neural network, comprising an input layer module, a modulation layer module and an output layer module connected in sequence;

[0008] Wherein, the input layer module is used to generate incident light beams of different wavelengths;

[0009] The modulation layer module is used to perform phase modulation on the incident light beam generated by the input layer module, and encrypt the information to be transmitted which is split into multiple target images through optical diffraction calculation. The modulation layer module includes multiple modulation layers connected in sequence, and the switching state of each modulation layer is independently controllable. The trained modulation layer module can generate a unique corresponding key for each target image, and each key contains the wavelength of the incident light beam and the switching state of each modulation layer;

[0010] The output layer module is used to reproduce the split information to be transmitted and realize optical secret sharing. The output layer module includes multiple output planes, and the distance from each output plane to the last modulation layer is different.

[0011] Further, the type of the incident light beam includes a plane wave.

[0012] Furthermore, the modulation layer is constructed based on a cascaded diffractive optical element.

[0013] Furthermore, the modulation layer is constructed based on a spatial light modulator.

[0014] Furthermore, the modulation layer module is trained by using a back propagation algorithm in combination with a stochastic gradient descent algorithm until the gradient converges to a local optimum.

[0015] Furthermore, in the back-propagation algorithm, a loss function is constructed based on the mean square error between the target image reproduced by the output layer module and the real target image.

[0016] Furthermore, in the modulation layer module, each modulation layer includes a plurality of neurons, each neuron is regarded as a secondary source of the incident light beam, and information transmission is achieved between the modulation layers through the diffraction of light.

[0017] Furthermore, the manner in which information is transmitted between modulation layers is determined by the transfer coefficient.

[0018] Furthermore, the transfer coefficient includes an amplitude term and a phase term.

[0019] The present invention also provides a multi-wavelength optical secret sharing method based on an optical diffraction neural network, comprising the following steps:

[0020] Obtain the information to be transmitted and split it into multiple target images;

[0021] generating incident light beams of different wavelengths;

[0022] Construct and train a modulation layer module, which includes multiple modulation layers connected in sequence. The switching state of each modulation layer is independently controllable. The incident light beam is phase modulated by the trained modulation layer module, and the split information to be transmitted is encrypted by optical diffraction calculation. Each target image corresponds to a key, and each key contains the wavelength of the incident light beam and the switching state of each modulation layer.

[0023] The corresponding target image is reproduced according to each key to achieve optical secret sharing.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] The present invention designs a multi-wavelength optical secret sharing system based on an optical diffraction neural network, comprising an input layer module, a modulation layer module and an output layer module connected in sequence, wherein the input layer module is used to generate incident light beams of different wavelengths, the modulation layer module is used to perform phase modulation on the incident light beam, and encrypt the information to be transmitted which is split into multiple target images through optical diffraction calculation, the modulation layer module comprises multiple modulation layers connected in sequence, the switching state of each modulation layer is independently controllable, and the trained modulation layer module can generate a unique corresponding key for each target image, each key contains the wavelength of the incident light beam and the switching state of each modulation layer, and the security of information transmission is further improved by increasing the types of information contained in the key; the output layer module comprises multiple output planes, and the distances from each output plane to the last modulation layer are different, which provides new possibilities for expanding the encrypted information from two dimensions to three dimensions, and can further improve the capacity of information transmission. Only when all keys are accurate and complete can complete information be obtained, and the security of information transmission is further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a schematic diagram of the structure of the system of the present invention,

[0027] Description of reference numerals: 1, input layer module, 2, modulation layer module, 3, output layer module;

[0028] Figure 2 Different keys and their corresponding decrypted images in the embodiment;

[0029] Figure 3 is the quality assessment result of the output image (with mean square error MSE as an indicator);

[0030] Figure 4 The quality evaluation result of the output image (with peak signal-to-noise ratio PSNR as an indicator);

[0031] Figure 5 The functional test results of the embodiment are as follows:

[0032] Among them, (5a) and (5b) are the test results of two target images in the first set of data, (5c) and (5d) are the test results of two target images in the second set of data, (5e) and (5f) are the test results of two target images in the third set of data, and (5g) and (5h) are the test results of two target images in the fourth set of data. DETAILED DESCRIPTION

[0033] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0034] Example 1

[0035] This embodiment provides a multi-wavelength optical secret sharing system based on optical diffraction neural network, such as Figure 1 As shown, it includes an input layer module 1, a modulation layer module 2 and an output layer module 3 which are connected in sequence. The detailed functions of each module are as follows:

[0036] A. Input layer module 1

[0037] The input layer module 1 is used to generate incident light beams of different wavelengths. The types of incident light beams include but are not limited to plane waves and Gaussian beams. The wavelength of the incident light beam can be changed by changing the parameter λ i size achieved.

[0038] B. Modulation layer module 2

[0039] The modulation layer module 2 is used to phase modulate the incident light beam generated by the input layer module 1, and encrypt the information to be transmitted which is split into multiple target images through optical diffraction calculation. Figure 1 As shown, in this embodiment, the modulation layer module 2 is an optical diffraction neural network including multiple modulation layers, and the transmission between adjacent modulation layers is realized by the diffraction of light. Each modulation layer contains multiple neurons, and each neuron is regarded as a secondary source of the incident light beam. The input of the neuron of the next modulation layer is defined as the result of the output of all neurons in the previous layer being superimposed on the neuron after diffraction propagation. The way in which the i-th neuron in the l-th modulation layer transmits the received information is determined by the transfer coefficient Determine the transfer coefficient Including amplitude terms and the phase term In this embodiment, only the phase change is considered, so the amplitude term is regarded as a constant 1.

[0040] The on / off state of each modulation layer is independently controllable. The trained modulation layer module 2 can generate a unique key for each target image. Each key is a set of wavelengths of the incident light beam and the on / off state of each modulation layer. The modulation layer can be realized by cascading diffractive optical elements (DOE) or spatial light modulators (SLM).

[0041] C. Output layer module 3

[0042] The output layer module 3 is used to reproduce the split information to be transmitted. In this embodiment, the output layer module 3 includes multiple output planes, each output plane has a different distance to the last modulation layer, and each output plane is used to reproduce a corresponding target image according to a key.

[0043] Using the back propagation algorithm and the stochastic gradient descent algorithm, M groups of data are selected from the data set (each group of images contains N target images), and the modulation layer module 2 is iteratively trained until the gradient converges to the local optimum. Finally, a set of phase terms of the modulation layer can be obtained. Parameters, in the back propagation algorithm, the loss function is constructed based on the mean square error between the target image reproduced by the output layer module 3 and the real target image. After the training is completed, the intensity distribution of the output light of the last modulation layer is close to the target image. The data sets used for training include but are not limited to the MNIST data set and the EMNIST data set.

[0044] To verify the effectiveness of the above system, this example conducts experiments on the MNIST dataset:

[0045] Three plane waves of different wavelengths with an intensity of 1 are selected as the incident light beams. The wavelengths are 635nm (red light), 532nm (green light) and 435nm (blue light), and the size of the incident area is 600*800 pixels. The number of modulation layers is 4, and the number of pixels in each layer is 600*800. The output layer has 2 output planes with pixels of 600*800. The distance between each modulation layer, between the first modulation layer and the input module, and between the output planes is 10cm. The two output planes are 0.1m and 0.2m away from the last modulation layer, respectively. Four groups of data are selected from the MNIST dataset, each group includes 2 target images. Each target image corresponds to a set of incident wave wavelengths and the switching state of the modulation layer. The specific corresponding relationships are as follows. Figure 2 shown.

[0046] Considering that the same modulation layer module has phase differences in the regulation of different wavelengths, the trained phase can be scaled according to the wavelength of the incident light beam. The system first forward propagates the incident light beam, and then uses the back propagation algorithm combined with the stochastic gradient descent algorithm to adjust the phase term of the modulation layer. The forward propagation algorithm is implemented by FFT (Fast Fourier Transform), and the back propagation algorithm is implemented by IFFT (Inverse Fast Fourier Transform). Add random phases to the target field to reduce crosstalk between output images. Calculate the error gradient of the loss function with respect to the phase parameter, and use the Adam optimization algorithm to update the phase term. Until the loss function converges, where the parameters β1 = 0.9, β2 = 0.999, ε = 10 -8 In this embodiment, 60 iterations are performed, and 10 batches of training are performed in each iteration.

[0047] In order to evaluate the quality of the output image, this embodiment selects the mean square error MSE and the peak signal-to-noise ratio PSNR as the judgment criteria. The experimental results are as follows: Figure 3 and Figure 4 As shown in the figure, it can be seen that as the number of iterations increases, the MSE and PSNR curves tend to be stable, with MSE of 1.075 and PSNR of 59.896 dB. The system was tested using the corresponding incident wave wavelength and the switching state of the modulation layer. The test results of the four sets of data are shown in the figure below. Figure 5 The encrypted information can be read only when the key is complete and correct.

[0048] The above system improves the security and information capacity of the optical secret sharing system by increasing the types of key information, and the output plane is expanded from a single plane to multiple planes, providing new possibilities for expanding the encrypted information in the optical secret sharing system from two dimensions to three dimensions.

[0049] Example 2

[0050] This embodiment provides a multi-wavelength optical secret sharing method based on an optical diffraction neural network, comprising the following steps:

[0051] S1. Obtain the information to be transmitted and split it into multiple target images.

[0052] S2. Generate incident light beams of different wavelengths.

[0053] S3. Construct and train a modulation layer module, wherein the modulation layer module includes multiple modulation layers connected in sequence, and the switching state of each modulation layer is independently controllable. The incident light beam is phase modulated by the trained modulation layer module, and the split information to be transmitted is encrypted through optical diffraction calculation. Each target image corresponds to a key, and each key contains the wavelength of the incident light beam and the switching state of each modulation layer.

[0054] S4. Reproduce the corresponding target image according to each key to achieve optical secret sharing.

[0055] If the above method is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0056] The above description of the embodiments is to facilitate the understanding and use of the invention by those skilled in the art. It is obvious that those skilled in the art can easily make various modifications to these embodiments and apply the general principles described herein to other embodiments without creative work. Therefore, the present invention is not limited to the above embodiments, and improvements and modifications made by those skilled in the art based on the disclosure of the present invention without departing from the scope of the present invention should be within the scope of protection of the present invention.

Claims

1. A multi-wavelength optical secret sharing system based on optical diffraction neural network, characterized in that: It comprises an input layer module (1), a modulation layer module (2) and an output layer module (3) which are connected in sequence; Wherein, the input layer module (1) is used to generate incident light beams of different wavelengths; The modulation layer module (2) is used to perform phase modulation on the incident light beam generated by the input layer module (1), and encrypt the information to be transmitted which is split into multiple target images through optical diffraction calculation. The modulation layer module (2) includes multiple modulation layers connected in sequence, and the switching state of each modulation layer is independently controllable. The trained modulation layer module (2) can generate a unique corresponding key for each target image, and each key contains the wavelength of the incident light beam and the switching state of each modulation layer. The output layer module (3) is used to reproduce the split information to be transmitted and realize optical secret sharing. The output layer module (3) comprises a plurality of output planes, and the distances from each output plane to the last modulation layer are different.

2. The multi-wavelength optical secret sharing system based on optical diffraction neural network according to claim 1, characterized in that: The type of the incident light beam includes a plane wave.

3. The multi-wavelength optical secret sharing system based on optical diffraction neural network according to claim 1, characterized in that: The modulation layer is constructed based on cascaded diffractive optical elements.

4. The multi-wavelength optical secret sharing system based on optical diffraction neural network according to claim 1, characterized in that: The modulation layer is constructed based on a spatial light modulator.

5. The multi-wavelength optical secret sharing system based on optical diffraction neural network according to claim 1, characterized in that: The modulation layer module (2) is trained by using a back propagation algorithm combined with a stochastic gradient descent algorithm until the gradient converges to a local optimum.

6. The multi-wavelength optical secret sharing system based on optical diffraction neural network according to claim 5, characterized in that: In the back propagation algorithm, a loss function is constructed based on the mean square error between the target image reproduced by the output layer module (3) and the real target image.

7. The multi-wavelength optical secret sharing system based on optical diffraction neural network according to claim 1, characterized in that: In the modulation layer module (2), each modulation layer includes a plurality of neurons, each neuron is regarded as a secondary source of an incident light beam, and information transmission is achieved between the modulation layers through the diffraction of light.

8. The multi-wavelength optical secret sharing system based on optical diffraction neural network according to claim 7, characterized in that: The way information is transmitted between the modulation layers is determined by the transfer coefficient.

9. The multi-wavelength optical secret sharing system based on optical diffraction neural network according to claim 8, characterized in that: The transfer coefficient includes an amplitude term and a phase term.

10. A multi-wavelength optical secret sharing method based on optical diffraction neural network, characterized in that: The following steps are involved: Obtain the information to be transmitted and split it into multiple target images; generating incident light beams of different wavelengths; Construct and train a modulation layer module, which includes multiple modulation layers connected in sequence. The switching state of each modulation layer is independently controllable. The incident light beam is phase modulated by the trained modulation layer module, and the split information to be transmitted is encrypted by optical diffraction calculation. Each target image corresponds to a key, and each key contains the wavelength of the incident light beam and the switching state of each modulation layer. The corresponding target image is reproduced according to each key to achieve optical secret sharing.

Images