Multidimensional encryption communication device based on optoelectronic neuromorphic function and implementation method thereof

Abstract

The application discloses a kind of multi-dimensional encryption communication devices based on optoelectronic neuro morphic function and implementation method thereof, it is related to hardware security communication technical field, core is in: the character information to be encrypted transmission is converted into binary sequence, according to predefined mapping rule, binary sequence is segmented and mapped as the wavelength, duration and quantity of optical pulse, and parameterized optical pulse sequence is generated;Optoelectronic neuro morphic device array is used to receive optical pulse sequence and generate unique corresponding electrical response signal, and the electrical response signal is the ciphertext after encryption;Pretrained convolutional neural network is used to identify and decrypt electrical response signal, and restore original character information.The application integrates information perception, physical encryption and neuro morphic decryption in one, with ultra-low power consumption, high security, reconfigurable, portable advantage, suitable for internet of things terminal, mobile device and maritime, aviation and other fields of secure communication.

Description

Technical Field

[0001] This invention relates to the field of hardware secure communication technology, and in particular to a multidimensional encrypted communication device based on optoelectronic neuromorphic functions and its implementation method. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] In recent years, the rapid development of IoT, edge computing, and artificial intelligence technologies has brought increasingly serious issues to information transmission security. Traditional information encryption methods mainly rely on software algorithms to protect data through complex mathematical transformations (such as asymmetric encryption and hash functions). However, when running on traditional von Neumann computing architectures, there are significant "power wall" and "memory wall" problems. Furthermore, the improvement and development of computing power, especially quantum computing power, threatens its long-term security.

[0004] Hardware security technologies based on novel nanodevices (such as memristors) have become a research hotspot due to their advantages such as physical non-cloning and low power consumption. Currently, there are methods that utilize the randomness of the formation and breakage of memristor conductive filaments to formulate encryption mechanisms. However, encryption mechanisms based on randomness are often sensitive to environmental factors (such as temperature and humidity) and circuit noise, and their stability and reliability are insufficient. They often require additional complex correction circuits, increasing system complexity and power consumption.

[0005] As an emerging in-memory computing device, optoelectronic neuromorphic devices can simultaneously respond to electrical and optical signals, simulating the dynamic behavior of biological synapses and neurons, making them an ideal platform for realizing low-power neuromorphic computing and visual perception. Existing research largely focuses on using these devices to simulate various synaptic plasticities to construct artificial neural networks, lacking applications in novel hardware encryption systems. Summary of the Invention

[0006] To address the shortcomings of the existing technologies, this invention provides a multidimensional encrypted communication device based on opto-neuromorphic functions and its implementation method. Considering the physical characteristics of opto-neuromorphic devices in responding to multidimensional light stimuli with deterministic and differentiated responses, information is encoded in the multidimensional parameters of light pulses. Stable, efficient, and highly secure hardware encryption is achieved through device physical conversion. Combined with neural network decryption, a complete encrypted communication chain that ensures security, low power consumption, and reconfigurability from a physical perspective is constructed, which can adapt to dynamic and complex actual communication environments.

[0007] In a first aspect, the present invention provides a multidimensional encrypted communication device based on optoelectronic neuromorphic functions.

[0008] A multidimensional encrypted communication device based on optoelectronic neuromorphic functions includes: The information encoding module is used to convert the character information to be encrypted into a binary sequence, and to segment the binary sequence into optical pulse wavelengths, durations and quantities according to predefined mapping rules, thereby generating a parameterized optical pulse sequence. The encryption execution module includes an array of optoelectronic neuromorphic devices for receiving a sequence of light pulses and generating a unique corresponding electrical response signal, which is the encrypted ciphertext. The decryption and recognition module is used to receive electrical response signals, and to use a pre-trained convolutional neural network to recognize and decrypt the electrical response signals to restore the original character information. The microcontroller integration platform is used to carry and run the decryption and identification module, enabling real-time decryption and information display of encrypted signals on mobile devices.

[0009] Further technical solutions also include: The optional two-factor authentication module includes a dynamic visual signal recognition unit based on reservoir calculation and a key verification unit based on device response, which is used to pre-authenticate the sender's identity and initiate the multi-dimensional encrypted communication process after successful verification.

[0010] A further technical solution is that the optoelectronic neuromorphic device array is based on PET / ITO / HfAlO x / NbO x The ITO structure, when stimulated by a sequence of light pulses with set wavelength, pulse width and number, simulates various plastic behaviors of biological synapses, generating repetitive and uniquely corresponding excitatory postsynaptic current responses, i.e. electrical response signals. Among these, various plastic behaviors include: excitatory postsynaptic currents, pulse number-dependent plasticity, pulse frequency-dependent plasticity, pulse width-dependent plasticity, and the transition from short-term plasticity to long-term plasticity.

[0011] A further technical solution is that the predefined mapping rule is: The 6-bit ASCII code representing a character is divided into three segments: the first two segments are mapped to the wavelength of the optical pulse, the middle two segments are mapped to the duration of a single pulse, and the last two segments are mapped to the total number of pulses.

[0012] In a further technical solution, the wavelength of the optical pulse is selected from multiple discrete wavelength values ​​ranging from the ultraviolet light band to the visible light band.

[0013] A further technical solution is that the convolutional neural network includes an input layer for receiving one-dimensional current waveform time-series data, multiple convolutional and pooling layers for extracting local spatiotemporal features of the current waveform, a fully connected layer for integrating all features, and an output layer with 26 neurons. The received electrical response signal is input into a pre-trained convolutional neural network, including: The encrypted one-dimensional current waveform timing data is received through the input layer and used as the original input data. Local spatiotemporal features of the current waveform are extracted by convolutional layers to capture key differences in the encrypted data; among these, local spatiotemporal features include peak value and relaxation rate. Dimensionality reduction of the output features of the convolutional layer is achieved through pooling layers; By integrating all deep features after pooling through a fully connected layer, a mapping relationship between deep features and character categories is established. By analyzing the category and recognition probability of the English letters corresponding to the output signal from the output layer, the signal decryption result is generated, and the original character information is restored.

[0014] A further technical solution is that the pre-training process of the convolutional neural network is as follows: For each English letter, the encrypted transmission experiment was repeated according to the predefined mapping rules, and a number of current response signals were collected as labeled positive samples to form a 26-class character dataset. Build a convolutional neural network model, initialize model parameters, and configure the loss function and optimizer; The dataset is divided into a training set and a validation set. The data is input into the model to calculate the prediction results through forward propagation. The weights are updated through the back propagation algorithm to minimize the prediction error. After each iteration of training, the accuracy of the model's recognition is evaluated using a validation set until the preset conditions are met, thus completing the training. The pre-trained model weight parameters are ported to the microcontroller integration platform to adapt to the real-time decryption requirements of mobile devices.

[0015] Secondly, the present invention provides a method for implementing a multidimensional encrypted communication device based on optoelectronic neuromorphic functions.

[0016] A method for implementing a multidimensional encrypted communication device based on optoelectronic neuromorphic functions, comprising: The character information to be encrypted and transmitted is converted into a binary sequence. According to the predefined mapping rules, the binary sequence is segmented and mapped to the wavelength, duration and number of optical pulses to generate a parameterized optical pulse sequence. A sequence of light pulses is input into an array of optoelectronic neuromorphic devices to generate a unique corresponding electrical response signal, which is then transmitted; the electrical response signal is the encrypted ciphertext. The system receives electrical response signals, uses a pre-trained convolutional neural network to identify and decrypt the signals, recovers the original character information, and completes encrypted communication.

[0017] Further technical solutions also include a two-factor authentication process before encrypted communication, as follows: When the sender initiates communication, it simultaneously sends dynamic flag signal actions and hides the steady-state current value key of the encryption device in the action features; the flag signal actions include: gestures and dynamic facial expressions; The receiver acquires dynamic flag signaling video streams through visual sensors, uses a storage pool to calculate and identify flag signaling trajectory in real time, completes preliminary identity verification, and simultaneously extracts hidden current value keys from the high-dimensional state space of the trajectory, comparing the extracted keys with locally stored legitimate keys. If the flag signal gestures are correctly identified and the key matches, pre-authentication is passed and subsequent encrypted communication is initiated; otherwise, encrypted communication is rejected.

[0018] Thirdly, the present invention also provides an application of the above-described device in secure communication of Internet of Things devices, encrypted information transmission of mobile terminals, and confidential communication systems in the maritime or aviation fields.

[0019] The above one or more technical solutions have the following beneficial effects: 1. This invention provides a multidimensional encrypted communication device based on optoelectronic neuromorphic function and its implementation method. Considering the physical characteristics of optoelectronic neuromorphic devices to multidimensional light stimuli such as wavelength, pulse width, and frequency, a corresponding encryption mechanism is designed to encode information in the multidimensional parameters of the light pulse. Stable, efficient, and highly secure hardware encryption is achieved by utilizing the physical conversion of the device. Combined with neural network decryption, a complete encrypted communication chain that is secure, low-power, and reconfigurable from the physical level is constructed, which can adapt to dynamic and complex actual communication environments.

[0020] 2. The encryption mechanism designed in this invention is rooted in the unique material physics properties and nonlinear response of the device, and does not rely on algorithms that can be broken by mathematical analysis, providing security independent of algorithms. The single-event energy consumption of the core encryption unit, namely the optoelectronic device, can be as low as 3.3 fJ, which is far lower than that of traditional encryption circuits, making it particularly suitable for energy-constrained edge devices. The encryption key space is composed of multiple physical dimensions such as wavelength, time, and pulse number. The entire encryption system can be flexibly and cost-effectively reset by redefining the mapping rules or changing the decryption neural network model. In addition, a single device can simultaneously complete optical signal sensing, neuromorphic information processing, and non-volatile / volatile state storage, enabling on-sensor computation and greatly reducing data transfer overhead.

[0021] 3. By integrating dynamic environment perception and identity pre-authentication mechanisms, this invention enables the system to cope with the practical application challenges of uncertain receiver identities and changing environments, such as in maritime communications, and realizes the evolution from static encryption devices to dynamic secure communication systems.

[0022] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0023] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0024] Figure 1 This is a schematic diagram of the overall architecture of multidimensional encrypted communication in Embodiment 1 of the present invention; Figure 2 The PET / ITO / HfAlO in Embodiment 1 of the present invention x / NbO x A schematic diagram of the fabrication of an ITO structured neuromorphic sensor; Figure 3 The PET / ITO / HfAlO in Embodiment 1 of the present invention x / NbO x XPS spectra of the ITO structured neuromorphic sensor material; where (a) represents Hf, (b) represents Al, and (c) represents Nb. Figure 4 This is a response curve of the neuromorphic sensor in Embodiment 1 of the present invention to light of different wavelengths; Figure 5 This is a schematic diagram of the internal changes of the neuromorphic device in Embodiment 1 of the present invention; Figure 6 This is a response curve of the device in Embodiment 1 of the present invention to two consecutive optical pulses; Figure 7 This is a response curve of the device in Embodiment 1 of the present invention under different numbers of light pulses. Figure 8 This is a response curve of the device in Embodiment 1 of the present invention to optical pulses of different pulse durations; Figure 9 The diagram shows the response curves of the device in Embodiment 1 of the present invention to optical pulses of different frequencies. Figure 10 This is a response curve diagram of the device in Embodiment 1 of the present invention when it is converted from STM to LTM; Figure 11 This is a schematic diagram of the corresponding table for encoding binary information into optical pulse information in Embodiment 1 of the present invention; where (a) is wavelength, (b) is pulse width, and (c) is quantity; Figure 12 This is a waveform diagram of the response of the device in Embodiment 1 of the present invention to the letter "S" under specific encoded optical pulse conditions; Figure 13This is a schematic diagram of the convolutional neural network used for decryption in Embodiment 1 of the present invention; Figure 14 This is a graph showing the success rate of identifying legitimate recipients and attackers in Embodiment 1 of the present invention.

[0025] Figure 15 This is a schematic diagram of the hardware architecture of the multi-dimensional encrypted communication device system in Embodiment 1 of the present invention. Detailed Implementation

[0026] It should be noted that the following detailed descriptions are exemplary and are intended only to describe specific embodiments and to provide further explanation of the invention, and are not intended to limit the scope of exemplary embodiments of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0027] Example 1 To address the problems existing in current software encryption and traditional hardware encryption technologies, this embodiment provides a multidimensional encrypted communication device based on optoelectronic neuromorphic functions, which specifically includes: The information encoding module is used to convert the character information to be encrypted into a binary sequence, and to segment the binary sequence into optical pulse wavelengths, durations and quantities according to predefined mapping rules, thereby generating a parameterized optical pulse sequence. The encryption execution module includes an array of optoelectronic neuromorphic devices for receiving a sequence of light pulses and generating a unique corresponding electrical response signal, which is the encrypted ciphertext. The decryption and recognition module is used to receive electrical response signals, and to use a pre-trained convolutional neural network to recognize and decrypt the electrical response signals to restore the original character information. The microcontroller integration platform is used to carry and run the decryption and identification module, enabling real-time decryption and information display of encrypted signals on mobile devices; The optional two-factor authentication module includes a dynamic visual signal recognition unit based on reservoir calculation and a key verification unit based on device response, which is used to pre-authenticate the sender's identity and initiate the multi-dimensional encrypted communication process after successful verification.

[0028] The following content provides a more detailed description of the multidimensional encrypted communication device based on optoelectronic neuromorphic function proposed in this embodiment.

[0029] like Figure 1As shown, the entire device system is a hardware and software collaborative platform integrating information sensing, physical encryption, neuromorphic decryption, and identity authentication. Its core idea lies in utilizing the physical characteristics of optoelectronic neuromorphic devices to produce deterministic and unique electrical responses to multi-parameter optical stimuli, constructing an end-to-end secure closed loop from information to physical signals and back to information. The workflow of this device system is as follows: First, based on the information encoding module (also known as the optical excitation module), this module is responsible for converting the digital character information (such as English letters) to be encrypted into an ASCII binary sequence. Then, according to the preset "wavelength-time-quantity" three-dimensional mapping rule, the binary sequence is segmented into an optical pulse sequence defined by three dimensions: wavelength, pulse duration, and pulse quantity. Based on the specific optical pulse parameters after the conversion, the encryption execution module can be driven.

[0030] Secondly, the encryption execution module, as the core hardware, is based on PET / ITO / HfAlO. x / NbO x An array of optoelectronic neuromorphic devices based on ITO material is constructed. This array receives the aforementioned light pulse sequence and generates a unique corresponding electrical response signal, thereby achieving physical-level information encryption. Specifically, under stimulation by light pulses of specific wavelengths, pulse widths, and numbers, the optoelectronic neuromorphic device array can generate highly repetitive and uniquely corresponding excitatory postsynaptic current responses. This response signal is the ciphertext encrypted using hardware.

[0031] Specifically, the core of the encryption execution module is a memristor device with optoelectronic neuromorphic characteristics. In this embodiment, a PET / ITO / HfAlO structure is used. x / NbO x An array of ITO stacked optoelectronic neuromorphic devices is described below, and the fabrication process and procedure are as follows: Figure 2 As shown, HfAlO was deposited on a PET / ITO substrate using atomic layer deposition (ALD) technology in a reaction chamber at 200°C. x Composite oxide thin films serve as the core functional layer of the device; in the deposited HfAlO x NbO was sequentially deposited on the thin film using a magnetron sputtering process. x The fabrication of the two-end devices is completed by depositing a layer and an ITO top electrode. PET (polyethylene terephthalate) is used as the bottom substrate, ITO is the bottom electrode layer, and HfAlO is deposited on top of it. x Hafnium aluminum oxide (HfAlO) is a material with a high dielectric constant and can be used as a gate dielectric material in transistors. x NbO deposited on thin film x Niobium oxide is a metal oxide with photoelectric properties.

[0032] To confirm the chemical composition and chemical state of the device material, X-ray photoelectron spectroscopy analysis was performed on the prepared thin film. The results are as follows: Figure 3 As shown, the XPS spectrum (X-ray photoelectron spectroscopy, where the horizontal axis represents binding energy and the vertical axis represents intensity) confirms the presence of elements such as Hf, Al, Nb, and O and their expected chemical states, ensuring the quality and consistency of the functional layer material.

[0033] The fabricated optoelectronic devices, under light pulse modulation, successfully simulated key plastic behaviors of various biological synapses, which is the physical basis for their ability to perform information encryption and processing. The multi-wavelength response characteristics of neuromorphic devices (or sensors) are as follows... Figure 4 As shown, it can produce significant excitatory postsynaptic current responses to light pulses from the ultraviolet to the visible light band (such as 310 nm, 450 nm, and 550 nm), and the response amplitude and relaxation dynamics exhibit distinguishable systematic differences due to different wavelengths. This provides a physical possibility for using wavelength as a encryption dimension. The mechanism for the different corresponding currents caused by different light pulses may originate from the difference in the strength of the conductive filaments inside the device, such as... Figure 5 As shown, it displays conductive filaments of different strengths, with the resistivity of the devices shown in (a) to (c) gradually decreasing.

[0034] In addition, the device exhibits rich synaptic dynamic behaviors, such as Figure 6 The demonstrated dual-pulse facilitation property, namely the higher EPSC amplitude induced by the second pulse in response to two consecutive light pulses, simulates the short-term enhancement effect of biological synapses.

[0035] like Figure 7 As shown, the device exhibits pulse number-dependent plasticity; with increasing stimulus pulse number, the EPSC amplitude shows a cumulative enhancement in a nonlinear response. Furthermore, the device's response characteristics also change systematically with variations in pulse duration and applied frequency, as demonstrated by the following: Figure 8 The pulse width dependence shown and as Figure 9 The pulse frequency-dependent plasticity is shown.

[0036] Under continuous light stimulation of a specific intensity or frequency, the device can achieve a state transition from short-term memory to long-term memory, such as... Figure 10 As shown, this is manifested by a significant increase in the current decay time constant, a characteristic analogous to the memory consolidation process in biological learning. More importantly, tests show that the device consumes as little as 3.3 fJ per synaptic event, an ultra-low power characteristic that lays a solid foundation for building extremely low-power encryption hardware suitable for edge devices.

[0037] Furthermore, the core of the multi-dimensional encrypted communication method proposed in this embodiment lies in mapping digital information onto multiple physical parameters of an optical pulse, and then performing physical obfuscation and conversion through the nonlinear response of the device. The mapping rules are as follows: Figure 11 As shown, taking the character "S" as an example, firstly, the character "S" to be transmitted is converted into its standard 6-bit ASCII code, i.e., the binary sequence "010011". Then, a multi-dimensional parameter mapping is performed, dividing this 6-bit binary code into three segments, with each pair of bits corresponding to a physical parameter: the first two bits "01" map to the light pulse wavelength (λ), corresponding to a wavelength of 450 nanometers according to a preset lookup table; the middle two bits "00" map to the single pulse duration (D), corresponding to a pulse width of 100 milliseconds; the last two bits "11" map to the total number of pulses (N), corresponding to the application of 30 consecutive pulses. Next, a programmable light source system generates a regular light pulse sequence with a wavelength of 450 nanometers, each pulse width of 100 milliseconds, and a total of 30 pulses. Then, the above light pulse sequence is irradiated onto a photoelectric neuromorphic device, where charge carriers inside the device are excited, transported, and trapped, generating a unique, nonlinear transient current response curve. This current waveform is the physical ciphertext of the character "S", as shown below. Figure 12 As shown.

[0038] The encryption process described above is entirely determined by the material physical properties of the device and the parameters of the optical pulse, making it difficult to derive from an external mathematical model. Finally, a high-precision data acquisition device is used to record the complete waveform of the current changing over time at both ends of the device. For all 26 letters of the English alphabet, a unique mapping relationship of "binary code → (λ, D, N) parameter set → characteristic current waveform" can be established according to this rule, forming a complete encryption dictionary.

[0039] Subsequently, the decryption and recognition module includes a signal acquisition and transmission unit responsible for acquiring the analog current waveform output by the device. The module also contains a pre-trained convolutional neural network that learns and memorizes the characteristics of legitimate encrypted electrical signals, enabling it to accurately reconstruct the original characters while maintaining an extremely low recognition rate against unauthorized attempts to crack the signal. Specifically, the pre-trained convolutional neural network intelligently identifies and classifies the acquired encrypted electrical signals (i.e., electrical response signals), accurately reconstructing the original character information. Preferably, to adapt to mobile and edge computing scenarios, the trained decryption convolutional neural network is deployed on an embedded microcontroller, forming a microcontroller integration platform (also known as a microcontroller integration and display platform), enabling portable real-time decryption and information display.

[0040] Specifically, due to the complexity and device-specific nature of the current waveforms generated by encryption, this embodiment employs a convolutional neural network under a supervised learning paradigm as the sole effective decryption tool for the authorized receiver. The implementation process is as follows: First, a dataset is constructed. For each English letter, following the aforementioned encryption rules, multiple encryption transmission experiments are repeated on the device, collecting a large amount of current response waveform data as positive samples, thus forming a labeled dataset containing 26 letters. Next, the neural network is designed and trained, i.e., a convolutional neural network structure is constructed, model parameters are initialized, and the loss function and optimizer are configured. In this embodiment, the constructed convolutional neural network is as follows: Figure 13 As shown, it includes an input layer for receiving one-dimensional current waveform time-series data, multiple convolutional and pooling layers for extracting local spatiotemporal features of the current waveform, a fully connected layer for integrating all features, and an output layer with 26 neurons.

[0041] The process of inputting the received electrical response signal into a pre-trained convolutional neural network includes: The encrypted one-dimensional current waveform timing data is received through the input layer and used as the original input data. Local spatiotemporal features of the current waveform are extracted by convolutional layers to capture key differences in the encrypted data; among these, local spatiotemporal features include peak value and relaxation rate. Dimensionality reduction of the output features of the convolutional layer is achieved through pooling layers; By integrating all deep features after pooling through a fully connected layer, a mapping relationship between deep features and character categories is established. By analyzing the category and recognition probability of the English letters corresponding to the output signal from the output layer, the signal decryption result is generated, and the original character information is restored.

[0042] Then, the CNN is fully trained using the constructed dataset, which is to divide the dataset into a training set and a validation set, input the dataset into the model for forward propagation to calculate the prediction results, and continuously optimize and update the network weights through the backpropagation algorithm to minimize the prediction error. After each iteration of training, the accuracy of the model's recognition is evaluated using the validation set until the preset conditions are met, and the training is completed, so that it can finally learn to accurately identify the corresponding letter category from complex current waveforms.

[0043] Decryption performance evaluation shows that, after training, the CNN can achieve a decryption and recognition accuracy of up to 97.40% for encrypted signals held by legitimate users and generated by the correct devices and rules. Figure 14 As shown. For unauthorized attackers, due to the inability to obtain the correct physical response characteristics of the device or the "wavelength-time-quantity" mapping rules, the accuracy of their attempts to crack the encryption does not improve throughout the entire training period, averaging only about 2.88%, which is comparable to the probability of random guessing. This experimentally confirms the extremely high security and robustness of the encryption scheme.

[0044] Preferably, to promote the practical application of the device system, the weight parameters of the trained CNN model are ported to a resource-constrained microcontroller to adapt to the real-time decryption requirements of mobile devices and enable mobile decryption. When the microcontroller receives encrypted current waveform data, it can use the built-in weights to perform forward propagation calculations, quickly obtain the decryption result, and drive an external display screen to display it, thus fully demonstrating the feasibility of a portable secure communication terminal.

[0045] Furthermore, for complex communication scenarios such as maritime and aviation communication where the identity of the receiver is uncertain and the environment is dynamically complex, this embodiment integrates an optional dynamic identity pre-authentication module, namely a two-factor authentication module, on top of the basic encryption and decryption framework. This constructs a two-factor security enhancement mechanism that deeply integrates hardware and software. This module includes a dynamic visual signal recognition unit based on a reservoir calculation and a key verification unit based on device response. It can use the reservoir calculation to identify the sender's dynamic semaphore actions in real time and verify the hidden key encoded by the device's steady-state current value. Only communication parties that pass this "friend or foe identification" pre-authentication can initiate the subsequent multi-dimensional encrypted communication process, thereby building a solid front-end security defense in an open environment.

[0046] In this embodiment, the two-factor authentication module integrates dynamic visual signal recognition (such as semaphore) with key verification based on device response. This module uses a reservoir to process timing signals to achieve synchronous verification of the sender's identity and the hidden key, adding a pre-authentication security layer to the system's communication in an open environment.

[0047] The implementation process of the above two-factor authentication mechanism includes two consecutive and related verification phases: The first stage is dynamic semaphore recognition, aimed at preliminary identity verification. Specifically, the sender initiating the communication first transmits a standard dynamic semaphore motion signal sequence through body movements. The receiver acquires a continuous video stream of the dynamic semaphore motion using a visual sensor and utilizes Reservoir Computing (RC), an efficient recurrent neural network (RNN) model adept at handling time-series signals, to identify and decode the continuous trajectory of the semaphore motion in real time. The RC system (i.e., the Reservoir Computing network, a simplified and efficient training paradigm of RNN, whose core is a randomly initialized and usually fixed recurrent neural network layer containing a large number of sparsely connected neurons; this network layer can also be called a reservoir) nonlinearly maps continuous pose changes to a high-dimensional dynamic state space, thereby accurately identifying the character content expressed by the dynamic semaphore.

[0048] The second stage is hidden hardware key verification, which is synchronized with and deeply intertwined with the first stage, aiming to perform secondary key confirmation. While sending semaphore signals, the sender has encoded and hidden the hardware key information for this communication session—the steady-state current value generated by an opto-neuromorphic device corresponding to a specific letter—within subtle features of the semaphore movements. The receiver's RC system, while recognizing the semaphore movement trajectory, simultaneously extracts these hidden current values ​​from its high-dimensional state space. The final dual verification logic is as follows: the system compares the extracted hidden current value with the expected reference value generated by the legitimate key and securely stored locally on the receiver. Only communication requests that simultaneously meet the two strict conditions of "correct semaphore movement content recognition" and "successful matching of the extracted hidden hardware key" are considered trustworthy by the system. Requests that fail this pre-authentication are immediately rejected, effectively resisting unauthorized access and man-in-the-middle attacks from the very beginning of the communication link, greatly enhancing the overall security and reliability of the system in open environments.

[0049] It should be clarified that the above process mainly involves the identification and authentication of the receiver. The sender does not need to verify the receiver's legitimacy before sending information (an unauthorized receiver cannot understand / obtain the ciphertext, ensuring information transmission security). Instead, the sender confirms the receiver's ability to decrypt the ciphertext and provide a correct response, thus ultimately confirming the receiver as a trusted partner. Furthermore, since electrical information is a hidden feature embedded in the semaphore signals, its extraction also requires a corresponding key-ciphertext pair. Extracting hidden information from semaphore actions requires prior knowledge of a precise "encoding-mapping" algorithm. However, this algorithm is part of the overall system design and is a "key" pre-installed in the legitimate receiver's RC system, which external attackers cannot crack. Even if they could crack and extract the hidden information, it wouldn't be a directly usable key. It needs to be compared with the expected value generated by legitimate devices and stored locally on the receiver to obtain its actual meaning. Therefore, the above semaphore transmission process does not cause data leakage and effectively ensures data transmission security.

[0050] like Figure 15 As shown, it demonstrates the complete hardware and algorithm integration architecture of this device system, reflecting the vertical integration from physical devices to top-level applications. This integration is based on PET / ITO / HfAlO. x / NbO x The system includes an ITO-structured optoelectronic encryption array, a precision light source control system capable of outputting multi-wavelength light pulses, a high-bandwidth data acquisition card, an embedded microcontroller loaded with decryption neural network weights, and host computer software responsible for coordination, scheduling, and visualization interaction.

[0051] The core advantage of this system lies in its ultra-low power consumption. The energy consumption of the core encryption unit for a single event is only at the femtojoule level, making it highly suitable for scenarios with severe energy constraints, such as IoT sensor nodes and wearable devices. Its security mechanism is rooted in the physical characteristics of the device, providing fundamental security independent of algorithms. This provides a new hardware security solution resistant to mathematical analysis for mobile terminals, drone swarms, and ship-to-shore communication. In addition, the "wavelength-time-quantity" mapping rules and decryption neural network models on which the system relies can be redefined and retrained according to the security strategy. This reconfigurability allows the encryption key to be updated flexibly and at low cost, effectively addressing security challenges under long-term service, demonstrating strong practicality and broad application prospects.

[0052] In summary, this embodiment details the entire process of this system, from device fabrication and characterization, functional verification, to the implementation of multi-dimensional encryption and decryption algorithms, to the construction of security enhancement mechanisms for complex scenarios, and finally to system integration demonstration. This system creatively uses the physical response characteristics of optoelectronic neuromorphic devices as the foundation of information security, integrates multi-dimensional optical coding and artificial intelligence decryption technologies, and innovatively integrates dynamic biometric authentication, constructing a complete solution for a next-generation hardware secure communication device system that is highly adaptable, secure, and low-power.

[0053] Example 2 This embodiment provides a method for implementing a multidimensional encrypted communication device based on optoelectronic neuromorphic function proposed in Embodiment 1, specifically including the following steps: The character information to be encrypted and transmitted is converted into a binary sequence. According to the predefined mapping rules, the binary sequence is segmented and mapped to the wavelength, duration and number of optical pulses to generate a parameterized optical pulse sequence. A sequence of light pulses is input into an array of optoelectronic neuromorphic devices to generate a unique corresponding electrical response signal, which is then transmitted; the electrical response signal is the encrypted ciphertext. The system receives electrical response signals, uses a pre-trained convolutional neural network to identify and decrypt the signals, recovers the original character information, and completes encrypted communication.

[0054] Furthermore, it also includes a two-factor authentication process before encrypted communication, as follows: When the sender initiates communication, it simultaneously sends dynamic flag signal actions and hides the steady-state current value key of the encryption device in the action features; the flag signal actions include: gestures and dynamic facial expressions; The receiver acquires dynamic flag signaling video streams through visual sensors, uses a storage pool to calculate and identify flag signaling trajectory in real time, completes preliminary identity verification, and simultaneously extracts hidden current value keys from the high-dimensional state space of the trajectory, comparing the extracted keys with locally stored legitimate keys. If the flag signal gestures are correctly identified and the key matches, pre-authentication is passed and subsequent encrypted communication is initiated; otherwise, encrypted communication is rejected.

[0055] Example 3 This embodiment provides an application of the device described in Embodiment 1 above in secure communication of Internet of Things devices, encrypted information transmission of mobile terminals, and confidential communication systems in the maritime or aviation fields.

[0056] The steps involved in Examples 2 and 3 above correspond to those in Example 1. For specific implementation details, please refer to the relevant description section of Example 1.

[0057] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0058] The above description is only a preferred embodiment of the present invention. Although the specific implementation of the present invention has been described in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that, based on the technical solution of the present invention, various modifications or variations that can be made by those skilled in the art without creative effort are still within the scope of protection of the present invention.

Claims

1. A multidimensional encrypted communication device based on optoelectronic neuromorphic function, characterized in that, include: The information encoding module is used to convert the character information to be encrypted into a binary sequence, and to segment the binary sequence into optical pulse wavelengths, durations and quantities according to predefined mapping rules, thereby generating a parameterized optical pulse sequence. The encryption execution module includes an array of optoelectronic neuromorphic devices for receiving a sequence of light pulses and generating a unique corresponding electrical response signal, which is the encrypted ciphertext. The decryption and recognition module is used to receive electrical response signals, and to use a pre-trained convolutional neural network to recognize and decrypt the electrical response signals to restore the original character information. The microcontroller integration platform is used to carry and run the decryption and identification module, enabling real-time decryption and information display of encrypted signals on mobile devices.

2. The multidimensional encrypted communication device based on optoelectronic neuromorphic function as described in claim 1, characterized in that, Also includes: The optional two-factor authentication module includes a dynamic visual signal recognition unit based on reservoir calculation and a key verification unit based on device response, which is used to pre-authenticate the sender's identity and initiate the multi-dimensional encrypted communication process after successful verification.

3. The multidimensional encrypted communication device based on optoelectronic neuromorphic function as described in claim 1, characterized in that, The optoelectronic neuromorphic device array is based on PET / ITO / HfAlO x / NbO x The ITO structure, when stimulated by a sequence of light pulses with set wavelength, pulse width and number, simulates various plastic behaviors of biological synapses, generating repetitive and uniquely corresponding excitatory postsynaptic current responses, i.e. electrical response signals. Among these, various plastic behaviors include: excitatory postsynaptic currents, pulse number-dependent plasticity, pulse frequency-dependent plasticity, pulse width-dependent plasticity, and the transition from short-term plasticity to long-term plasticity.

4. The multidimensional encrypted communication device based on optoelectronic neuromorphic function as described in claim 1, characterized in that, The predefined mapping rule is as follows: The 6-bit ASCII code representing a character is divided into three segments: the first two segments are mapped to the wavelength of the optical pulse, the middle two segments are mapped to the duration of a single pulse, and the last two segments are mapped to the total number of pulses.

5. The multidimensional encrypted communication device based on optoelectronic neuromorphic function as described in claim 4, characterized in that, The wavelength of the optical pulse is selected from multiple discrete wavelength values ​​ranging from the ultraviolet light band to the visible light band.

6. The multidimensional encrypted communication device based on optoelectronic neuromorphic function as described in claim 1, characterized in that, The convolutional neural network includes an input layer for receiving one-dimensional current waveform time-series data, multiple convolutional and pooling layers for extracting local spatiotemporal features of the current waveform, a fully connected layer for integrating all features, and an output layer with 26 neurons. The received electrical response signal is input into a pre-trained convolutional neural network, including: The encrypted one-dimensional current waveform timing data is received through the input layer and used as the original input data. Local spatiotemporal features of the current waveform are extracted by convolutional layers to capture key differences in the encrypted data; among these, local spatiotemporal features include peak value and relaxation rate. Dimensionality reduction of the output features of the convolutional layer is achieved through pooling layers; By integrating all deep features after pooling through a fully connected layer, a mapping relationship between deep features and character categories is established. By analyzing the category and recognition probability of the English letters corresponding to the output signal from the output layer, the signal decryption result is generated, and the original character information is restored.

7. The multidimensional encrypted communication device based on optoelectronic neuromorphic function as described in claim 6, characterized in that, The pre-training process of the convolutional neural network is as follows: For each English letter, the encrypted transmission experiment was repeated according to the predefined mapping rules, and a number of current response signals were collected as labeled positive samples to form a 26-class character dataset. Build a convolutional neural network model, initialize model parameters, and configure the loss function and optimizer; The dataset is divided into a training set and a validation set. The data is input into the model to calculate the prediction results through forward propagation. The weights are updated through the back propagation algorithm to minimize the prediction error. After each iteration of training, the accuracy of the model's recognition is evaluated using a validation set until the preset conditions are met, thus completing the training. The pre-trained model weight parameters are ported to the microcontroller integration platform to adapt to the real-time decryption requirements of mobile devices.

8. A method for implementing a multidimensional encrypted communication device based on optoelectronic neuromorphic functions, characterized in that, include: The character information to be encrypted and transmitted is converted into a binary sequence. According to the predefined mapping rules, the binary sequence is segmented and mapped to the wavelength, duration and number of optical pulses to generate a parameterized optical pulse sequence. A sequence of light pulses is input into an array of optoelectronic neuromorphic devices to generate a unique corresponding electrical response signal, which is then transmitted; the electrical response signal is the encrypted ciphertext. The system receives electrical response signals, uses a pre-trained convolutional neural network to identify and decrypt the signals, recovers the original character information, and completes encrypted communication.

9. The method for implementing the multidimensional encrypted communication device based on optoelectronic neuromorphic function as described in claim 8, characterized in that, It also includes a two-factor authentication process before encrypted communication, as follows: When the sender initiates communication, it simultaneously sends dynamic flag signal actions and hides the steady-state current value key of the encryption device in the action characteristics; The flag signaling actions include: hand gestures and facial expressions; The receiver acquires dynamic flag signaling video streams through visual sensors, uses a storage pool to calculate and identify flag signaling trajectory in real time, completes preliminary identity verification, and simultaneously extracts hidden current value keys from the high-dimensional state space of the trajectory, comparing the extracted keys with locally stored legitimate keys. If the flag signal gestures are correctly identified and the key matches, pre-authentication is passed and subsequent encrypted communication is initiated; otherwise, encrypted communication is rejected.

10. The application of a multidimensional encrypted communication device based on optoelectronic neuromorphic function as described in any one of claims 1-7 in secure communication of Internet of Things devices, encrypted information transmission of mobile terminals, and confidential communication systems in the maritime or aviation fields.

Images