High-speed data encryption and decryption system and method based on photon neural network

Abstract

The invention provides a high-speed data encryption and decryption system and method based on a photon neural network, and the system comprises a photon neural network encryption module which is used for receiving an incident light signal, and directly mapping the incident light signal into a noise-like encrypted light field in an optical domain; the transmission channel is used for transmitting the noise-like encrypted light field to a photon neural network decryption module; and the photon neural network decryption module is used for carrying out inverse transformation on the noise-like encrypted light field and restoring to obtain an emergent light signal consistent with the original information carried by the incident light signal. An encryption key is dynamically generated by adopting the trainable photon neural network, traditional fixed physical mask or electronic algorithm encryption is replaced, and meanwhile, high-speed, parallel and low-power-consumption data encryption transmission is realized through multi-dimensional optical characteristic integration and optical domain end-to-end processing.

Description

Technical Field

[0001] This invention relates to the fields of silicon photonic devices and optical neural networks, and in particular to a high-speed data encryption and decryption system and method based on photonic neural networks. Background Technology

[0002] With the development of information technology, traditional electronic encryption algorithms (such as AES and RSA) face speed bottlenecks and high energy consumption problems in high-speed optical communication systems. Optical encryption technology, utilizing the parallelism and multi-dimensional characteristics of light, offers a potential solution. Early optical encryption was based on methods such as random phase masks and double random phase encoding, but most relied on fixed physical keys, which would completely destroy security once obtained by an attacker.

[0003] In recent years, photonic neural networks (PNNs) have attracted widespread attention as an important development direction in the field of optical computing. Their basic principle is to utilize the interference, diffraction, and nonlinear effects of light to directly complete the core computation process of the neural network in the optical domain, avoiding the delay and energy consumption caused by photoelectric conversion. Compared to traditional electronic neural networks, photonic neural networks possess inherent parallel processing capabilities and extremely high data throughput, enabling simultaneous information encoding and computation using multiple dimensions of physical quantities such as the phase, amplitude, and polarization of light. This multi-dimensional joint control capability allows optical networks to process multiple encrypted information streams in parallel within the same optical path, significantly improving the system's information capacity and security. Simultaneously, by flexibly configuring phase and polarization states, dynamically reconfigurable encryption mapping functions can be achieved, further enhancing the system's resistance to attacks. With the increasing integration of optical devices and advancements in control technology, PNNs are gradually evolving from single-function to reconfigurable and multi-functional applications, demonstrating increasing potential in areas such as image recognition, signal processing, and optical communication security. Especially in the field of optical encryption, PNN-based systems can directly implement encryption mapping and feature transformation in the optical domain, providing a new technical path for achieving high-speed, low-power physical layer secure communication. Summary of the Invention

[0004] This invention proposes a high-speed data encryption and decryption system and method based on photonic neural networks to solve or partially solve the above-mentioned problems.

[0005] One aspect of the present invention provides a high-speed data encryption and decryption system based on a photonic neural network, the system comprising: A photonic neural network encryption module, comprising a first multi-layer cascaded diffraction structure, wherein the optical parameters of the first multi-layer cascaded diffraction structure can be adjusted by a dynamic encryption key, wherein the photonic neural network encryption module is used to receive incident light signals and directly map the incident light signals in the optical domain into a noise-like encrypted light field; A transmission channel is used to transmit the noise-like encrypted light field to the photonic neural network decryption module; The photonic neural network decryption module includes a second multi-layer cascaded diffraction structure. The optical parameters of the second multi-layer cascaded diffraction structure can be adjusted by a dynamic decryption key that uniquely matches the dynamic encryption key. The photonic neural network decryption module is used to perform an inverse transformation on the noise-like encrypted light field to restore the output light signal that is consistent with the original information carried by the incident light signal.

[0006] Furthermore, the optical parameters include at least one of the phase, polarization, wavelength, amplitude, and rotation angle of the light field; the dynamic encryption key and the dynamic decryption key are respectively the encryption configuration value of the photonic neural network encryption module and the decryption configuration value of the photonic neural network decryption module obtained by a preset joint optimization algorithm.

[0007] Furthermore, both the first and second multi-layer cascaded diffraction structures are integrated planar waveguide structures, fabricated on quartz wafers using electron beam lithography and plasma etching processes.

[0008] Furthermore, the photonic neural network encryption module and the photonic neural network decryption module respectively integrate a laser diode array and a Mach-Zehnder modulator; the laser diode array and the Mach-Zehnder modulator are respectively integrated with the first multilayer cascaded diffraction structure and the second multilayer cascaded diffraction structure through flip-chip bonding process.

[0009] Furthermore, both the first and second multi-layer cascaded diffraction structures are internally configured with nested structures formed by micro-ring resonators and special waveguides.

[0010] Furthermore, the system also includes: The optical emission module is used to convert the input electrical signal into an incident optical signal that carries the original information; An optical receiving module is used to convert the emitted optical signal into an electrical signal that restores the original information.

[0011] Another aspect of the present invention provides an encryption and decryption method based on the above-described high-speed data encryption and decryption system based on photonic neural networks, the method comprising: The photonic neural network encryption module and the photonic neural network decryption module are jointly trained based on a joint optimization algorithm to obtain a unique mapping relationship between the dynamic encryption key and the dynamic decryption key; Based on the unique mapping relationship between the dynamic encryption key and the dynamic decryption key, the encryption configuration parameters of the photonic neural network encryption module and the decryption configuration parameters of the photonic neural network decryption module are adjusted respectively to restore the incident light signal to the output light signal that retains the original information after encryption, transmission and decryption.

[0012] Furthermore, the joint training of the photonic neural network encryption module and the photonic neural network decryption module based on the joint optimization algorithm includes: S11. Establish the encryption light field transmission model of the photonic neural network encryption module and the decryption light field transmission model of the photonic neural network decryption module respectively. S12. Select plaintext-ciphertext training pairs from the preset training set, input the plaintext into the encrypted light field transmission model to generate an encrypted light field, and then input the encrypted light field into the decryption light field transmission model to obtain the decryption output result. S13. Using the bit error rate between the decryption output and the original plaintext as the core loss function, and combining it with the mean square error to construct a joint loss evaluation index; S14. Based on the joint loss evaluation index, the configuration parameters of the encrypted light field transmission model and the decrypted light field transmission model are adjusted synchronously using the error backpropagation algorithm. S15. Repeat S12-S14 until the mean square error is less than the preset error threshold and the bit error rate is less than the preset safety threshold, and the training converges. S16. Use the optimal configuration parameters of the encrypted optical field transmission model as the dynamic encryption key, use the configuration parameters of the decrypted optical field transmission model corresponding to the dynamic encryption key as the dynamic decryption key, and save the mapping relationship between the dynamic encryption key and the dynamic decryption key.

[0013] Furthermore, establishing the encryption light field transmission model of the photonic neural network encryption module and the decryption light field transmission model of the photonic neural network decryption module respectively includes: Based on the angular spectrum diffraction theory, the initial encryption light field transmission model of the photonic neural network encryption module and the initial decryption light field transmission model of the photonic neural network decryption module are established respectively. The actual structural parameters of the photonic neural network encryption module and the photonic neural network decryption module were collected through online calibration training. Based on the actual structural parameters, weight mapping models for the photonic neural network encryption module and the photonic neural network decryption module are constructed respectively. The initial encrypted light field transmission model is corrected by combining the error correction amount output by the weight mapping model of the photonic neural network encryption module to obtain the encrypted light field transmission model; The initial decryption light field transmission model is corrected by combining the error correction amount output by the weight mapping model of the photonic neural network decryption module to obtain the decryption light field transmission model.

[0014] This invention provides a high-speed data encryption and decryption system based on photonic neural networks. The system includes a photonic neural network encryption module, a transmission channel, and a photonic neural network decryption module. The photonic neural network encryption module enables multi-degree-of-freedom encryption and control of the incident light signal at the physical level, making it difficult to simulate decryption through software even if the encrypted light field is intercepted. Simultaneously, the photonic neural network encryption and decryption modules can encrypt and decrypt data within the optical domain, solving the photoelectric conversion bottleneck in traditional encryption systems and fully leveraging the high speed and low power consumption advantages of photonic processing. Furthermore, the dynamic encryption and decryption keys, with a unique mapping relationship, are adjusted in real-time and rapidly through electro-optic or thermo-optic effects, achieving a high-security "one-time key" mechanism.

[0015] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0016] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. In the drawings: Figure 1 This is a schematic diagram of the structure of a high-speed data encryption and decryption system based on a photonic neural network according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating the workflow of a high-speed data encryption and decryption system based on a photonic neural network, according to an embodiment of the present invention. Figure 3 A comparison image of the input image, encrypted image, and decrypted image of the high-speed data encryption and decryption system based on photonic neural networks provided in an embodiment of the present invention. Detailed Implementation

[0017] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0018] Introducing photonic neural networks into the encryption and decryption process not only allows for the direct construction of complex nonlinear encryption mappings in the optical domain, but also leverages multidimensional physical quantities such as the phase and polarization of light as natural key carriers, significantly enhancing the system's resistance to attacks and information concealment. Compared to traditional electronic encryption systems limited by analog-to-digital conversion bandwidth and serial processing architecture, the end-to-end processing approach in the optical domain achieves a better balance between real-time performance, energy efficiency, and security, effectively addressing the rate bottlenecks and energy consumption issues faced by electronic encryption schemes in high-speed optical communication scenarios. Therefore, employing photonic neural networks to construct optical encryption and decryption systems is an effective technological path to overcome existing technological limitations and meet the security requirements of next-generation optical communication.

[0019] See Figure 1 As shown, the high-speed data encryption and decryption system based on photonic neural networks disclosed in this embodiment of the invention specifically includes: The photonic neural network encryption module 101 includes a first multi-layer cascaded diffraction structure, and the optical parameters of the first multi-layer cascaded diffraction structure can be adjusted by a dynamic encryption key. The photonic neural network encryption module 101 is used to receive incident light signals and directly map the incident light signals into a noise-like encrypted light field in the optical domain. Transmission channel 102 is used to transmit the noise-like encrypted light field to photonic neural network decryption module 103; A photonic neural network decryption module 103 includes a second multi-layer cascaded diffraction structure. The optical parameters of the second multi-layer cascaded diffraction structure can be adjusted using a dynamic decryption key uniquely matched with the dynamic encryption key. The photonic neural network decryption module 103 is used to inversely transform the noise-like encrypted light field to restore the output light signal consistent with the original information carried by the incident light signal. Understandably, to facilitate the conversion of electrical signals into optical signals, the system provided in this embodiment may further include: The optical emission module is used to convert the input electrical signal into an incident optical signal that carries the original information; An optical receiving module is used to convert the emitted optical signal into an electrical signal that restores the original information.

[0020] The high-speed data encryption and decryption system based on photonic neural networks provided in this invention can encrypt or decrypt optical signals. The photonic neural network encryption module 101 can generate an encrypted light field through multi-degree-of-freedom control. After transmission through a channel, the light field is inversely transformed by the decryption network to recover the original information. The programmable photonic neural network module is an integrated photonic chip / module that uses optical signals for neural network computation and is dynamically programmable / reconfigurable in hardware. The weights, connections, and activations of the artificial neural network are implemented using light interference, diffraction, and phase / intensity modulation. Furthermore, through electro-optical programming, the network structure and weights can be reconfigured without replacing the chip, thereby enabling dynamic encryption and decryption key transformations. Network parameters can be periodically updated according to security policies, achieving advanced "one-time key" security protection.

[0021] Furthermore, in this embodiment of the invention, the optical parameters include at least one of the phase, polarization, wavelength, amplitude, and rotation angle of the light field; the dynamic encryption key and the dynamic decryption key are respectively the encryption configuration value of the photonic neural network encryption module 101 and the decryption configuration value of the photonic neural network decryption module 103 obtained by calculating through a preset joint optimization algorithm.

[0022] Specifically, each diffraction structure in the photonic neural network module achieves complex linear and nonlinear transformations by adjusting parameters such as the phase, amplitude, or polarization of the light field. Its core advantage lies in its high parallelism and high-speed processing capabilities, making it very suitable for the high-speed computing needs in encryption tasks.

[0023] During the encryption process, the original data is mapped into a noisy encrypted light field by the photonic neural network encryption module 101, as shown in Formula 1: E out =F DNN ( E in ; θ (1) in E in and E out These are the input and output light fields of the photonic neural network encryption module 101, respectively. F DNN For the transformation function of the photonic neural network encryption module 101, θ It is a dynamic encryption key.

[0024] The decryption process is accomplished through a reverse diffraction network, the mathematical expression of which is shown in Formula 2: (2) in D out The output light field of the photonic neural network decryption module 103. This is the transformation function for the photonic neural network decryption module 103. This is a dynamic decryption key.

[0025] The unique advantage of the encryption mechanism in this invention lies in its dynamic key mechanism, which allows network parameters to be updated in real time through training optimization, achieving the high security requirement of "one-time key". At the same time, the physical layer security features ensure that the encryption effect is closely dependent on the actual optical path and component characteristics, making it difficult to simulate and reproduce through pure software. In addition, the multi-dimensional control capability, by combining multi-degree-of-freedom parameters such as polarization and wavelength, significantly expands the key space and effectively improves the system's defense against brute-force attacks.

[0026] Furthermore, this invention significantly enhances security by constructing a high-dimensional encryption space through the joint manipulation of multiple physical dimensions, including polarization, wavelength, and spatial modes. Specifically, anisotropic metasurfaces (such as birefringent nanopillars) are used to achieve polarization-sensitive optical field manipulation, ensuring that the decryption process requires matching the correct polarization key. Wavelength division multiplexing (WDM) technology is combined to split data into different wavelength channels for parallel processing, effectively increasing system capacity and cracking difficulty. Simultaneously, a multimode waveguide structure is employed, and the coupling relationship between modes is controlled through carefully designed diffraction patterns, further enhancing the complexity and uniqueness of the encryption mapping. This multi-degree-of-freedom collaborative mechanism gives the encryption optical field high-dimensional nonlinear characteristics, making it difficult for attackers to fully recover the key even if they obtain partial parameters, thus forming a multi-layered security protection system.

[0027] Furthermore, in this embodiment of the invention, both the first and second multi-layer cascaded diffraction structures are integrated planar waveguide structures, fabricated on a quartz wafer using electron beam lithography and plasma etching processes. This fabrication method can form a compact planar waveguide integrated diffraction optical neural network chip. This chip architecture integrates three diffraction layers on the same substrate, achieving high-precision alignment and a computational speed of up to TOPS, providing a hardware foundation for high-speed encryption.

[0028] Furthermore, the photonic neural network encryption module 101 and the photonic neural network decryption module 103 respectively integrate a laser diode array and a Mach-Zehnder modulator; the laser diode array and the Mach-Zehnder modulator are respectively integrated with the first multi-layer cascaded diffraction structure and the second multi-layer cascaded diffraction structure through flip-chip bonding process.

[0029] Furthermore, both the first and second multi-layer cascaded diffraction structures contain nested structures formed by micro-ring resonators and special waveguides. These nested structures enhance the interaction length between light and materials, thereby improving diffraction efficiency.

[0030] The embodiments of this invention employ a nested design of a micro-ring resonator and a special waveguide to enhance the interaction length between light and materials, thereby improving diffraction efficiency. During integration, active alignment technology is required to ensure that the relative positional accuracy of each optical element reaches the sub-micron level, which is crucial for maintaining the precise correspondence between the encryption and decryption networks.

[0031] Furthermore, embodiments of the present invention also provide an encryption and decryption method based on the above-described high-speed data encryption and decryption system using photonic neural networks, such as... Figure 2 As shown, the method includes: S1. Jointly train the photonic neural network encryption module 101 and the photonic neural network decryption module 103 based on a joint optimization algorithm to obtain a unique mapping relationship between the dynamic encryption key and the dynamic decryption key; S2. Based on the unique mapping relationship between the dynamic encryption key and the dynamic decryption key, adjust the encryption configuration parameters of the photonic neural network encryption module 101 and the decryption configuration parameters of the photonic neural network decryption module 103 respectively, so as to restore the incident light signal to the output light signal that retains the original information after encryption, transmission and decryption.

[0032] Furthermore, the joint training of the photonic neural network encryption module 101 and the photonic neural network decryption module 103 based on the joint optimization algorithm in step S1 includes: S11. Establish the encrypted light field transmission model of the photonic neural network encryption module 101 and the decrypted light field transmission model of the photonic neural network decryption module 103 respectively. S12. Select plaintext-ciphertext training pairs from the preset training set, input the plaintext into the encrypted light field transmission model to generate an encrypted light field, and then input the encrypted light field into the decryption light field transmission model to obtain the decryption output result. S13. Using the bit error rate between the decryption output and the original plaintext as the core loss function, and combining it with the mean square error to construct a joint loss evaluation index; S14. Based on the joint loss evaluation index, the configuration parameters of the encrypted light field transmission model and the decrypted light field transmission model are adjusted synchronously using the error backpropagation algorithm. S15. Repeat S12-S14 until the mean square error is less than the preset error threshold and the bit error rate is less than the preset safety threshold, and the training converges. S16. Use the optimal configuration parameters of the encrypted optical field transmission model as the dynamic encryption key, use the configuration parameters of the decrypted optical field transmission model corresponding to the dynamic encryption key as the dynamic decryption key, and save the mapping relationship between the dynamic encryption key and the dynamic decryption key.

[0033] Further, in step S11 of this embodiment of the invention, establishing the encrypted light field transmission model of the photonic neural network encryption module 101 and the decrypted light field transmission model of the photonic neural network decryption module 103 respectively includes: establishing the initial encrypted light field transmission model of the photonic neural network encryption module 101 and the initial decrypted light field transmission model of the photonic neural network decryption module 103 based on angular spectrum diffraction theory; collecting the actual structural parameters of the photonic neural network encryption module 101 and the photonic neural network decryption module 103 respectively through online calibration training; constructing the weight mapping model of the photonic neural network encryption module 101 and the photonic neural network decryption module 103 respectively based on the actual structural parameters; correcting the initial encrypted light field transmission model by combining the error correction amount output by the weight mapping model of the photonic neural network encryption module 101 to obtain the encrypted light field transmission model; and correcting the initial decrypted light field transmission model by combining the error correction amount output by the weight mapping model of the photonic neural network decryption module 103 to obtain the decrypted light field transmission model.

[0034] In this embodiment of the invention, the actual structural parameters include processing and assembly errors, diffraction layer spacing, refractive index of the core optical material, and waveguide loss coefficient. This embodiment of the invention can avoid transmission errors caused by hardware structure, significantly improving the robustness and prediction accuracy of the chip in actual operation. This method achieved 90.6% consistency and 2.3% arithmetic mean standard deviation in experiments, demonstrating good stability.

[0035] Furthermore, in this embodiment of the invention, adjusting the control parameters of the photonic neural network encryption module 101 and the photonic neural network decryption module 103 based on the unique mapping relationship between the dynamic encryption key and the dynamic decryption key specifically involves: calling the dynamic encryption key and adjusting the adjustable control parameters of the photonic neural network encryption module 101 to put the photonic neural network encryption module 101 into a preset encryption state; simultaneously calling the corresponding dynamic decryption key and adjusting the adjustable control parameters of the photonic neural network decryption module 103 to complete the key pairing between the decryption module and the encryption module.

[0036] In this embodiment of the invention, after the encryption module and decryption module are paired, encryption and decryption transmission can be performed. Specifically, this includes electro-optical conversion and encryption: the input electrical signal (such as a high-speed digital signal) to be transmitted is transmitted to the optical transmitting module, which converts the electrical signal into a continuous wavelength original optical signal and inputs it to the photonic neural network encryption module 101 loaded with a dynamic encryption key; through the coordinated modulation of the metasurface, polarization controller and other structures inside the encryption module, the original optical signal is converted into a noise-like encrypted optical field, which is transmitted at high speed through an optical transmission link (such as an optical fiber or a free-space optical link); decryption and electro-optical conversion: the noise-like encrypted optical field is transmitted to the photonic neural network decryption module 103 corresponding to the optical receiving module, which performs inverse diffraction modulation on the encrypted optical field based on the loaded dynamic decryption key to accurately restore the original optical signal; the optical receiving module converts the restored original optical signal into an electrical signal and outputs it to the terminal device to achieve high-fidelity restoration of the original information.

[0037] The photonic neural network of this invention achieves encryption and decryption based on optical field diffraction modulation. The transmission rate of the optical signal is not limited by the electron mobility. With the help of a precise optical field transmission model and error compensation mechanism, the data restoration accuracy is ensured in high-speed transmission scenarios (such as data transmission rates of 100 Gbps and above).

[0038] Based on existing technology, the system provided in this embodiment of the invention is expected to achieve the following performance: a single-channel data transmission rate of 100 Gbps, an eight-channel parallel transmission rate of 800 Gbps, and a system bit error rate of less than [missing information]. (That is, less than 1 error occurs in every trillion bits), the effective key space is greater than This combination effectively resists brute-force attacks. The system's encryption and decryption effects will be achieved through methods such as... Figure 3 The comparison diagrams shown intuitively demonstrate that the original input image becomes a speckle pattern resembling noise after being encrypted by a diffraction neural network. However, after being decrypted with the correct key, the original information can be recovered with high fidelity. The structural similarity index between the decrypted output and the original input image is expected to be higher than 0.95, which fully proves the reliability and security of the system.

[0039] The high-speed data encryption and decryption system and method based on photonic neural networks provided in this invention utilizes a programmable photonic chip to construct a diffraction neural network. Through the control of multi-degree-of-freedom optical parameters (including phase, polarization, wavelength, and spatial position), direct encryption and decryption of data in the optical domain is achieved. The core innovation lies in using a trainable photonic neural network to dynamically generate encryption keys, replacing traditional fixed physical masks or electronic algorithms. Simultaneously, through the integration of multi-dimensional optical characteristics and end-to-end optical domain processing, high-speed, parallel, and low-power encrypted data transmission is achieved.

[0040] Compared to traditional electronic encryption algorithms, this invention offers several orders of magnitude faster processing speed and possesses inherent resistance to quantum computing attacks. Compared to traditional optical encryption, the system features a dynamically reconfigurable encryption mechanism, significantly enhancing security. In scenarios such as fiber optic communication, data center optical interconnects, and secure communication networks, the system can achieve real-time, low-error-rate secure transmission, and its hardware architecture is based on existing optoelectronic device integration, demonstrating practicality and reliability.

[0041] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art and should not be interpreted in an idealized or overly formal sense unless specifically defined.

[0042] For the sake of simplicity, the method embodiments are described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0043] Furthermore, those skilled in the art will understand that although some embodiments herein include certain features included in other embodiments but not others, combinations of features from different embodiments are intended to be within the scope of the invention and form different embodiments. For example, any of the claimed embodiments can be used in any combination.

[0044] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A high-speed data encryption and decryption system based on photonic neural networks, characterized in that, The system includes: A photonic neural network encryption module, comprising a first multi-layer cascaded diffraction structure, wherein the optical parameters of the first multi-layer cascaded diffraction structure can be adjusted by a dynamic encryption key, wherein the photonic neural network encryption module is used to receive incident light signals and directly map the incident light signals in the optical domain into a noise-like encrypted light field; A transmission channel is used to transmit the noise-like encrypted light field to the photonic neural network decryption module; The photonic neural network decryption module includes a second multi-layer cascaded diffraction structure. The optical parameters of the second multi-layer cascaded diffraction structure can be adjusted by a dynamic decryption key that uniquely matches the dynamic encryption key. The photonic neural network decryption module is used to perform an inverse transformation on the noise-like encrypted light field to restore the output light signal that is consistent with the original information carried by the incident light signal.

2. The system according to claim 1, characterized in that, The optical parameters include at least one of the phase, polarization, wavelength, amplitude, and rotation angle of the light field; the dynamic encryption key and the dynamic decryption key are respectively the encryption configuration value of the photonic neural network encryption module and the decryption configuration value of the photonic neural network decryption module obtained by a preset joint optimization algorithm.

3. The system according to claim 2, characterized in that, Both the first and second multi-layer cascaded diffraction structures are integrated planar waveguide structures, fabricated on quartz wafers using electron beam lithography and plasma etching processes.

4. The system according to claim 3, characterized in that, The photonic neural network encryption module and the photonic neural network decryption module are respectively integrated with a laser diode array and a Mach-Zehnder modulator; the laser diode array and the Mach-Zehnder modulator are respectively integrated with the first multi-layer cascaded diffraction structure and the second multi-layer cascaded diffraction structure through flip-chip bonding process.

5. The system according to claim 4, characterized in that, Both the first and second multi-layer cascaded diffraction structures have nested structures formed by micro-ring resonators and special waveguides.

6. The system according to any one of claims 1-5, characterized in that, The system also includes: The optical emission module is used to convert the input electrical signal into an incident optical signal that carries the original information; An optical receiving module is used to convert the emitted optical signal into an electrical signal that restores the original information.

7. An encryption and decryption method for a high-speed data encryption and decryption system based on a photonic neural network as described in any one of claims 1-6, characterized in that, The method includes: The photonic neural network encryption module and the photonic neural network decryption module are jointly trained based on a joint optimization algorithm to obtain a unique mapping relationship between the dynamic encryption key and the dynamic decryption key; Based on the unique mapping relationship between the dynamic encryption key and the dynamic decryption key, the encryption configuration parameters of the photonic neural network encryption module and the decryption configuration parameters of the photonic neural network decryption module are adjusted respectively to restore the incident light signal to the output light signal that retains the original information after encryption, transmission and decryption.

8. The method according to claim 7, characterized in that, The joint training of the photonic neural network encryption module and the photonic neural network decryption module based on the joint optimization algorithm includes: S11. Establish the encryption light field transmission model of the photonic neural network encryption module and the decryption light field transmission model of the photonic neural network decryption module respectively. S12. Select plaintext-ciphertext training pairs from the preset training set, input the plaintext into the encrypted light field transmission model to generate an encrypted light field, and then input the encrypted light field into the decryption light field transmission model to obtain the decryption output result. S13. Using the bit error rate between the decryption output and the original plaintext as the core loss function, and combining it with the mean square error to construct a joint loss evaluation index; S14. Based on the joint loss evaluation index, the configuration parameters of the encrypted light field transmission model and the decrypted light field transmission model are adjusted synchronously using the error backpropagation algorithm. S15. Repeat S12-S14 until the mean square error is less than the preset error threshold and the bit error rate is less than the preset safety threshold, and the training converges. S16. Use the optimal configuration parameters of the encrypted optical field transmission model as the dynamic encryption key, use the configuration parameters of the decrypted optical field transmission model corresponding to the dynamic encryption key as the dynamic decryption key, and save the mapping relationship between the dynamic encryption key and the dynamic decryption key.

9. The method according to claim 8, characterized in that, The establishment of the encryption light field transmission model for the photonic neural network encryption module and the decryption light field transmission model for the photonic neural network decryption module includes: Based on the angular spectrum diffraction theory, the initial encryption light field transmission model of the photonic neural network encryption module and the initial decryption light field transmission model of the photonic neural network decryption module are established respectively. The actual structural parameters of the photonic neural network encryption module and the photonic neural network decryption module were collected through online calibration training. Based on the actual structural parameters, weight mapping models for the photonic neural network encryption module and the photonic neural network decryption module are constructed respectively. The initial encrypted light field transmission model is corrected by combining the error correction amount output by the weight mapping model of the photonic neural network encryption module to obtain the encrypted light field transmission model; The initial decryption light field transmission model is corrected by combining the error correction amount output by the weight mapping model of the photonic neural network decryption module to obtain the decryption light field transmission model.

Images