A VGA-oriented electromagnetic communication method and device

Abstract

The present invention discloses a VGA-oriented electromagnetic communication method and device, the method comprising: performing time-sharing processing on N electromagnetic communication controlled terminals in the same period and network, selecting the electromagnetic communication controlled terminal for communication; using the electromagnetic communication controlled terminal for communication to send data to the electromagnetic communication master terminal through the VGA interface of the electromagnetic communication controlled terminal; using the electromagnetic communication master terminal to restore the received data to achieve electromagnetic communication. The method of the present invention uses efficient electromagnetic-based isolated network covert access and communication control technology, combined with acoustic-based isolated network covert access and electromagnetic medium-based communication channel construction, to build a covert communication bridge between internal and external networks, obtain high-value information of the isolated internal network on demand, and then reversely achieve VGA-oriented electromagnetic communication.

Description

Technical Field

[0001] The present invention relates to the technical field of electromagnetic communication, and in particular to a VGA-oriented electromagnetic communication method and device. Background Art

[0002] The main function of electromagnetic radiation isolation network access is to modulate the data to be sent on the idle frequency band in the 88MHz to 108MHz FM frequency band (equivalent to the FM frequency band of general radio), and use the idle video output port (VGA) of the electromagnetic communication controlled end to send electromagnetic signals. Our personnel can use mobile terminals such as smart phones to receive and restore the data, establish a one-way data return channel in the isolation network, and can also cooperate with other means (such as sound waves) to establish a two-way data transmission channel.

[0003] Physical isolation refers to a network security measure that uses physical means to isolate network systems that need to avoid the risk of sensitive or critical information leakage from the Internet and other unsafe networks, such as military defense systems, command and control centers of critical infrastructure, financial systems, and government systems. Electromagnetic communication systems are a type of communication method that uses electromagnetic waves to transmit information. With the rapid development of science and technology, electromagnetic communication systems play an increasingly important role in important isolated networks such as military, infrastructure, finance, and government affairs. However, at the same time, the security issues of electromagnetic communication systems have become increasingly prominent. Electromagnetic communication systems have been attacked or destroyed, resulting in frequent incidents of serious consequences such as information leakage and system paralysis. Their security has a direct impact on national security, economic development, and social stability. How to ensure the security of electromagnetic communication systems and prevent information leakage and attacks has become a major issue that we must face and solve.

[0004] In order to cope with the global electromagnetic information security challenge, the security assessment of electromagnetic communication systems has become a research field that has attracted much attention. At present, domestic and foreign scholars have conducted a lot of research on electromagnetic communication systems and achieved certain results.

[0005] First, in terms of electromagnetic communication methods, researchers have proposed a variety of methods, such as model-based methods, risk-based methods, attack-based methods, etc. These methods have their own advantages and disadvantages, and need to be selected and optimized according to specific application scenarios and needs.

[0006] Secondly, in terms of electromagnetic communication security assessment technology, researchers have carried out research and application of various electromagnetic communication data security technologies, such as signal processing technology, encryption technology, intrusion detection technology, etc. These technologies provide strong support for electromagnetic communication security assessment. In addition, in terms of electromagnetic communication security assessment practice, researchers have carried out multiple practical projects, such as security testing of electromagnetic communication systems, vulnerability mining of electromagnetic communication systems, etc. These practical projects provide practical experience and technical accumulation for electromagnetic communication security assessment.

[0007] However, electromagnetic communication still faces some challenges and problems. First, the complexity of electromagnetic communication systems makes security assessment difficult. Second, existing methods and technologies still have certain limitations and need to be further improved. In addition, as new technologies continue to emerge, how to adapt to the development and application of new technologies is also an important issue facing electromagnetic communication. Summary of the invention

[0008] The technical problem to be solved by the present invention is to provide an electromagnetic communication method and device for VGA. Aiming at the characteristics of the target network adopting the physical isolation protection mechanism, which cannot be directly contacted through the network and the traditional security penetration testing means will fail, through the efficient electromagnetic-based isolated network covert access and communication control technology, combined with the acoustic-based isolated network covert access and electromagnetic medium-based communication channel construction, a covert communication bridge is built between the internal and external networks, and high-value information of the isolated internal network is obtained on demand, thereby realizing the reverse electromagnetic communication for VGA.

[0009] In order to solve the above technical problems, the first aspect of the embodiment of the present invention discloses a VGA-oriented electromagnetic communication method, the method comprising:

[0010] S1, performing time-sharing processing on N electromagnetic communication controlled terminals in the same period and network, and selecting the electromagnetic communication controlled terminal for communication, including:

[0011] S11, collecting information from the electromagnetic communication controlled terminals in the same period and network to obtain characteristic information of the electromagnetic communication controlled terminals;

[0012] The characteristic information includes attribute information and behavior information;

[0013] The attribute information includes network traffic information, system configuration information, hardware information, operating system version information, and application version information;

[0014] The behavior information includes communication flow information and communication return time information;

[0015] S12, preprocessing the feature information to obtain preprocessed feature information, including:

[0016] S121, performing data cleaning on the feature information to remove missing values ​​and outliers to obtain first feature information;

[0017] S122, constructing data indicators for the first characteristic information to obtain second characteristic information;

[0018] S123, performing data standardization processing on the second characteristic information to obtain third characteristic information;

[0019] S124, processing the third feature information to obtain pre-processed feature information;

[0020] S13, using the preprocessed feature information, training a preset data underlying distribution statistical model to obtain an optimized data underlying distribution statistical model;

[0021] S14, using the optimized data underlying distribution statistical model, predicting N electromagnetic communication controlled terminals in the same network at the same time period in the isolated network to be processed, and selecting an electromagnetic communication controlled terminal for communication;

[0022] The electromagnetic communication controlled terminal for communication includes a VGA interface;

[0023] S2, using the VGA interface of the electromagnetic communication controlled terminal for communication, sending data to the electromagnetic communication master terminal;

[0024] S3, using the electromagnetic communication main control terminal to restore the received data to achieve electromagnetic communication, including:

[0025] S31, using the electromagnetic communication main control terminal, performing denoising processing on the received data to obtain denoised data information;

[0026] S32, filtering the denoised data information to obtain filtered data information;

[0027] The signal filtering method is:

[0028] The signal to be processed is converted into the time-frequency domain using STFT, using a window length of 256 samples, an overlap of 75% and a Hamming window;

[0029] Reduce the length of the spectrum vector to 129 by discarding frequency samples corresponding to negative frequencies;

[0030] Define a fully convolutional network consisting of 16 convolutional layers. The first layer has a convolution kernel size of 9×8 and 18 convolution kernels. The second to 13th convolutional layers are a group of 3 layers repeated 4 times, with convolution kernel widths of 5, 9, and 9 respectively, and the number of convolution kernels is 64. The convolution kernel width of the last convolution layer is 129 and there is only 1 convolution kernel. In this network, convolution is performed only in the frequency dimension, and for all layers except the first layer, the convolution kernel width along the time dimension is set to 1. After the convolution layer, it passes through a batch normalization layer and a ReLU activation function layer, and finally outputs through a regression layer;

[0031] Establish a convolutional neural network to complete the mapping F between the STFT spectrum X(m, k) of the signal to be processed and the STFT spectrum S(m, k) of the clean signal to be processed, so as to achieve the purpose of denoising;

[0032] The loss function of the denoising network is expressed as:

[0033]

[0034] Among them, B represents the batch size, M and K represent the total number of STFT time frames and frequency points respectively; finally, the denoised signal to be processed is converted back to the time domain signal using the amplitude spectrum output by the network and the phase of the received signal to be processed;

[0035] S33, demodulating the filtered data information to obtain demodulated data information;

[0036] S34, decrypting the demodulated data information to obtain decrypted data information;

[0037] S35, decoding the decrypted data information to obtain decoded data information, and realizing electromagnetic communication.

[0038] A second aspect of an embodiment of the present invention discloses a VGA-oriented electromagnetic communication device, the device comprising:

[0039] The electromagnetic communication controlled terminal selection module is used to perform time-sharing processing on N electromagnetic communication controlled terminals in the same period and network, and select the electromagnetic communication controlled terminal for communication, including:

[0040] S11, collecting information from the electromagnetic communication controlled terminals in the same period and network to obtain characteristic information of the electromagnetic communication controlled terminals;

[0041] The characteristic information includes attribute information and behavior information;

[0042] The attribute information includes network traffic information, system configuration information, hardware information, operating system version information, and application version information;

[0043] The behavior information includes communication flow information and communication return time information;

[0044] S12, preprocessing the feature information to obtain preprocessed feature information, including:

[0045] S121, performing data cleaning on the feature information to remove missing values ​​and outliers to obtain first feature information;

[0046] S122, constructing data indicators for the first characteristic information to obtain second characteristic information;

[0047] S123, performing data standardization processing on the second characteristic information to obtain third characteristic information;

[0048] S124, processing the third feature information to obtain pre-processed feature information;

[0049] S13, using the preprocessed feature information, training a preset data underlying distribution statistical model to obtain an optimized data underlying distribution statistical model;

[0050] S14, using the optimized data underlying distribution statistical model, predicting N electromagnetic communication controlled terminals in the same network at the same time period in the isolated network to be processed, and selecting an electromagnetic communication controlled terminal for communication;

[0051] The electromagnetic communication controlled terminal for communication includes a VGA interface;

[0052] A data sending module, used to send data to the electromagnetic communication master terminal by using the VGA interface of the electromagnetic communication controlled terminal for communication;

[0053] The data restoration module is used to restore the received data using the electromagnetic communication main control terminal to realize electromagnetic communication, including:

[0054] S31, using the electromagnetic communication main control terminal, performing denoising processing on the received data to obtain denoised data information;

[0055] S32, filtering the denoised data information to obtain filtered data information;

[0056] The signal filtering method is:

[0057] The signal to be processed is converted into the time-frequency domain using STFT, using a window length of 256 samples, an overlap of 75% and a Hamming window;

[0058] Reduce the length of the spectrum vector to 129 by discarding frequency samples corresponding to negative frequencies;

[0059] Define a fully convolutional network consisting of 16 convolutional layers. The first layer has a convolution kernel size of 9×8 and 18 convolution kernels. The second to 13th convolutional layers are a group of 3 layers repeated 4 times, with convolution kernel widths of 5, 9, and 9 respectively, and the number of convolution kernels is 64. The convolution kernel width of the last convolution layer is 129 and there is only 1 convolution kernel. In this network, convolution is performed only in the frequency dimension, and for all layers except the first layer, the convolution kernel width along the time dimension is set to 1. After the convolution layer, it passes through a batch normalization layer and a ReLU activation function layer, and finally outputs through a regression layer;

[0060] Establish a convolutional neural network to complete the mapping F between the STFT spectrum X(m, k) of the signal to be processed and the STFT spectrum S(m, k) of the clean signal to be processed, so as to achieve the purpose of denoising;

[0061] The loss function of the denoising network is expressed as:

[0062]

[0063] Among them, B represents the batch size, M and K represent the total number of STFT time frames and frequency points respectively; finally, the denoised signal to be processed is converted back to the time domain signal using the amplitude spectrum output by the network and the phase of the received signal to be processed;

[0064] S33, demodulating the filtered data information to obtain demodulated data information;

[0065] S34, decrypting the demodulated data information to obtain decrypted data information;

[0066] S35, decoding the decrypted data information to obtain decoded data information, and realizing electromagnetic communication.

[0067] The third aspect of the present invention discloses another VGA-oriented electromagnetic communication device, the device comprising:

[0068] A memory storing executable program code;

[0069] a processor coupled to the memory;

[0070] The processor calls the executable program code stored in the memory to execute part or all of the steps in the VGA-oriented electromagnetic communication method disclosed in the first aspect of the embodiment of the present invention.

[0071] The fourth aspect of the present invention discloses a computer storable medium, which stores computer instructions. When the computer instructions are called, they are used to execute some or all of the steps in the VGA-oriented electromagnetic communication method disclosed in the first aspect of the embodiment of the present invention.

[0072] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0073] The inventive method is directed to the target network using a physical isolation protection mechanism, and cannot be directly contacted through the network, and the characteristics that traditional security penetration testing means will fail, through efficient electromagnetic-based isolated network concealed access and communication control technology, combined with acoustic-based isolated network concealed access, based on electromagnetic medium communication channel construction, a concealed communication bridge is constructed between the internal and external networks, and high-value information of the isolated intranet is obtained on demand, and then electromagnetic communication for VGA is reversely realized. The inventive method has strong usability, does not use conventional network connections (such as Ethernet, WiFi, mobile communications, etc.), and can realize isolated network unidirectional data transmission only through an idle VGA interface. With the popularization of high-definition video ports and the design of standard dual VGA interfaces, it is relatively common to have an idle VGA interface, so the system is highly practical. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0075] Figure 1 It is a flowchart of a VGA-oriented electromagnetic communication method disclosed in an embodiment of the present invention;

[0076] Figure 2 It is a flow chart of another VGA-oriented electromagnetic communication method disclosed in an embodiment of the present invention;

[0077] Figure 3 It is a schematic diagram of the deployment method of the electromagnetic communication system disclosed in the embodiment of the present invention;

[0078] Figure 4 It is a software module diagram of the electromagnetic communication system disclosed in the embodiment of the present invention;

[0079] Figure 5 It is a working diagram of the electromagnetic communication system disclosed in the embodiment of the present invention;

[0080] Figure 6 It is a structural schematic diagram of a VGA-oriented electromagnetic communication device disclosed in an embodiment of the present invention;

[0081] Figure 7 It is a schematic structural diagram of another VGA-oriented electromagnetic communication device disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0082] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0083] The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, device, product or equipment that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or equipment.

[0084] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0085] The present invention discloses a VGA-oriented electromagnetic communication method and device, the method comprising: performing time-sharing processing on N electromagnetic communication controlled terminals in the same period and the same network, selecting the electromagnetic communication controlled terminal for communication; using the electromagnetic communication controlled terminal for communication to send data to the electromagnetic communication master terminal through the VGA interface of the electromagnetic communication controlled terminal; using the electromagnetic communication master terminal to restore the received data to achieve electromagnetic communication. The method of the present invention uses efficient electromagnetic-based isolated network covert access and communication control technology, combined with acoustic-based isolated network covert access and electromagnetic medium-based communication channel construction, to build a covert communication bridge between internal and external networks, obtain high-value information of the isolated internal network on demand, and then reversely achieve VGA-oriented electromagnetic communication. The following is a detailed description of each.

[0086] Embodiment 1

[0087] See also Figure 1 , Figure 1 : is a flowchart of a VGA-oriented electromagnetic communication method disclosed in an embodiment of the present invention. Figure 1The VGA-oriented electromagnetic communication method described is used to establish a one-way data return channel in an isolation network, and can also be used in conjunction with other means, such as sound waves, to establish a two-way data transmission channel, which is not limited in the embodiments of the present invention. Figure 1 As shown, the VGA-oriented electromagnetic communication method may include the following operations:

[0088] S1, performing time-sharing processing on N electromagnetic communication controlled terminals in the same period and network, and selecting the electromagnetic communication controlled terminal for communication, including:

[0089] S11, collecting information from the electromagnetic communication controlled terminals in the same period and network to obtain characteristic information of the electromagnetic communication controlled terminals;

[0090] The characteristic information includes attribute information and behavior information;

[0091] The attribute information includes network traffic information, system configuration information, hardware information, operating system version information, and application version information;

[0092] The behavior information includes communication flow information and communication return time information;

[0093] S12, preprocessing the feature information to obtain preprocessed feature information, including:

[0094] S121, performing data cleaning on the feature information to remove missing values ​​and outliers to obtain first feature information;

[0095] S122, constructing data indicators for the first characteristic information to obtain second characteristic information;

[0096] S123, performing data standardization processing on the second characteristic information to obtain third characteristic information;

[0097] S124, processing the third feature information to obtain pre-processed feature information;

[0098] S13, using the preprocessed feature information, training a preset data underlying distribution statistical model to obtain an optimized data underlying distribution statistical model;

[0099] S14, using the optimized data underlying distribution statistical model, predicting N electromagnetic communication controlled terminals in the same network at the same time period in the isolated network to be processed, and selecting an electromagnetic communication controlled terminal for communication;

[0100] The electromagnetic communication controlled terminal for communication includes a VGA interface;

[0101] S2, using the VGA interface of the electromagnetic communication controlled terminal for communication, sending data to the electromagnetic communication master terminal;

[0102] S3, using the electromagnetic communication main control terminal to restore the received data to achieve electromagnetic communication, including:

[0103] S31, using the electromagnetic communication main control terminal, performing denoising processing on the received data to obtain denoised data information;

[0104] S32, filtering the denoised data information to obtain filtered data information;

[0105] The signal filtering method is:

[0106] The signal to be processed is converted into the time-frequency domain using STFT, using a window length of 256 samples, an overlap of 75% and a Hamming window;

[0107] Reduce the length of the spectrum vector to 129 by discarding frequency samples corresponding to negative frequencies;

[0108] Define a fully convolutional network consisting of 16 convolutional layers. The first layer has a convolution kernel size of 9×8 and 18 convolution kernels. The second to 13th convolutional layers are a group of 3 layers repeated 4 times, with convolution kernel widths of 5, 9, and 9 respectively, and the number of convolution kernels is 64. The convolution kernel width of the last convolution layer is 129 and there is only 1 convolution kernel. In this network, convolution is performed only in the frequency dimension, and for all layers except the first layer, the convolution kernel width along the time dimension is set to 1. After the convolution layer, it passes through a batch normalization layer and a ReLU activation function layer, and finally outputs through a regression layer;

[0109] Establish a convolutional neural network to complete the mapping F between the STFT spectrum X(m, k) of the signal to be processed and the STFT spectrum S(m, k) of the clean signal to be processed, so as to achieve the purpose of denoising;

[0110] The loss function of the denoising network is expressed as:

[0111]

[0112] Among them, B represents the batch size, M and K represent the total number of STFT time frames and frequency points respectively; finally, the denoised signal to be processed is converted back to the time domain signal using the amplitude spectrum output by the network and the phase of the received signal to be processed;

[0113] S33, demodulating the filtered data information to obtain demodulated data information;

[0114] S34, decrypting the demodulated data information to obtain decrypted data information;

[0115] S35, decoding the decrypted data information to obtain decoded data information, and realizing electromagnetic communication.

[0116] It can be seen that the method of the present invention is aimed at the target network adopting the physical isolation protection mechanism, which cannot be directly contacted through the network and the characteristics that the traditional security penetration test means will fail. Through efficient electromagnetic-based isolated network concealed access and communication control technology, combined with acoustic-based isolated network concealed access and electromagnetic medium-based communication channel construction, a concealed communication bridge is built between the internal and external networks, and high-value information of the isolated intranet is obtained on demand, and then electromagnetic communication for VGA is realized in reverse. The method of the present invention has strong usability, does not use conventional network connections (such as Ethernet, WiFi, mobile communications, etc.), and can realize unidirectional data transmission of the isolated network only through an idle VGA interface. The method of the present invention is highly practical. With the popularization of high-definition video ports and the design of standard dual VGA interfaces, it is more common to have an idle VGA interface, so the system is highly practical.

[0117] Embodiment 2

[0118] See also Figure 2 , Figure 2 FIG. 1 is a flow chart of another VGA-oriented electromagnetic communication method disclosed in an embodiment of the present invention. Figure 2 The VGA-oriented electromagnetic communication method described is used to establish a one-way data return channel in an isolation network, and can also be used in conjunction with other means, such as sound waves, to establish a two-way data transmission channel, which is not limited in the embodiments of the present invention. Figure 2 As shown, the VGA-oriented electromagnetic communication method may include the following operations:

[0119] The electromagnetic communication main control terminal program includes two modules: data receiving module and data parsing module. The data receiving module mainly receives FM band electromagnetic radiation signals through mobile terminal devices such as mobile phones, and improves the signal-to-noise ratio of the received signal through signal filtering technology, and demodulates, decodes and decrypts the signal to restore the original data or file. Figure 3 It is a schematic diagram of the deployment method of the electromagnetic communication system disclosed in the embodiment of the present invention; Figure 4 It is a software module diagram of the electromagnetic communication system disclosed in the embodiment of the present invention;

[0120] The signal filtering method is wavelet denoising:

[0121] 1) Perform wavelet transform on noisy signals: decompose the signal into various scales.

[0122] a) Define wavelet basis function: Select a suitable wavelet basis function, such as Haar, Daubechies, Symmlet, Coiflet, etc. This embodiment uses the Daubechies (dbN) wavelet function (N is the order of the wavelet).

[0123] b) Signal decomposition: The signal to be processed is decomposed into multiple wavelet coefficients through fast wavelet transform (FWT) or multi-resolution analysis (MRA), and these coefficients correspond to wavelet components of different frequencies.

[0124] The specific process of fast wavelet transform (FWT) is as follows:

[0125] The signal is sampled and quantized to discretize the continuous signal.

[0126] Perform wavelet transform on discrete signals and decompose the signals into multiple wavelet coefficients.

[0127] The wavelet coefficients are processed, such as threshold processing, smoothing processing, etc., to remove noise or perform signal compression.

[0128] The processed wavelet coefficients are inversely transformed to obtain the processed signal.

[0129] The specific process of multi-resolution analysis (MRA) is as follows:

[0130] Select an appropriate filter (i.e. wavelet basis) to filter the signal.

[0131] The filtered signal is sampled and quantized to obtain sub-signals at multiple scales.

[0132] Analyze and process the sub-signals at each scale to extract useful information.

[0133] The sub-signals at various scales are synthesized to obtain signal representations at multiple resolutions.

[0134] 2) Threshold quantization of high-frequency coefficients: At each scale, a threshold is selected and the high-frequency coefficients of this layer are soft-thresholded. Soft thresholding is a commonly used threshold processing method that sets coefficients greater than the threshold to zero and keeps coefficients less than the threshold unchanged.

[0135] Determine the threshold: Mainly based on the adaptive threshold method, the threshold is determined according to the local variance of each wavelet coefficient. For each wavelet coefficient, its local variance is calculated, and then an adaptive threshold is determined based on this variance. This method can better handle wavelet coefficients of different scales and different positions.

[0136] Soft thresholding process:

[0137] a) In the calculation process of the adaptive threshold method, a not too large positive number is selected as the threshold, otherwise the output features will all be zero, and the denoising will lose its meaning.

[0138] b) Perform wavelet transform on the input signal to obtain a set of wavelet coefficients.

[0139] c) Apply a soft thresholding function to each wavelet coefficient, that is, compare the wavelet coefficient with the threshold. If the absolute value of the wavelet coefficient is less than the threshold, the wavelet coefficient is set to 0; if the absolute value of the wavelet coefficient is greater than the threshold, the wavelet coefficient is shrunk to near 0.

[0140] d) Perform inverse transform on the processed wavelet coefficients to obtain the denoised signal.

[0141] Hard thresholding and soft thresholding are both threshold processing methods in wavelet transform, and they differ in processing methods and effects. Hard thresholding is a simple and direct method that sets wavelet coefficients less than the threshold to 0 and retains wavelet coefficients greater than the threshold. This processing method can effectively remove noise, but may cause loss of sample details. Soft thresholding is a smoother processing method that shrinks wavelet coefficients less than the threshold to zero, while retaining wavelet coefficients greater than the threshold. This processing method can better preserve the sample's detailed information, but may introduce some new noise.

[0142] Soft Thresholding Function

[0143]

[0144] In order to remove noise during the soft thresholding process, further processing is required:

[0145] Soft thresholding can remove noise by following these steps:

[0146] a) Perform wavelet decomposition on the signal to obtain wavelet coefficients at each scale.

[0147] b) Perform soft thresholding on the wavelet coefficients, that is, shrink the wavelet coefficients that are less than the threshold to zero.

[0148] c) Perform an inverse transform on the processed wavelet coefficients to obtain the denoised signal. (The principle of soft thresholding is to determine whether it is noise based on the amplitude of the wavelet coefficients. If the amplitude of the wavelet coefficients is small, it is considered to be noise and shrinks to zero; if the amplitude of the wavelet coefficients is large, it is considered to be a signal and remains unchanged. In this way, noise can be effectively removed and the detailed information of the signal can be retained.)

[0149] 3) Wavelet reconstruction: Based on the low-frequency sparseness of the Nth layer of wavelet decomposition and the modified high-frequency coefficients of the first layer to the Nth layer, the wavelet reconstruction of the two-dimensional signal is calculated. Wavelet reconstruction specifically includes:

[0150] a) Perform inverse transform on the decomposed wavelet coefficients to obtain sub-signals at various scales.

[0151] b) Synthesize the sub-signals at each scale to obtain an approximate reconstruction of the original signal.

[0152] It should be noted that the process of wavelet reconstruction is to synthesize the sub-signals obtained by decomposition to obtain the original signal. Specifically, the low-frequency components and high-frequency components can be upsampled, and then convolved with the wavelet basis function, and finally the results are added to obtain an approximate reconstruction of the original signal. By continuously performing such reconstruction, the details of the original signal can be gradually restored.

[0153] 4) Inverse transform of wavelet coefficients

[0154] First, the wavelet coefficients are inversely transformed to obtain sub-signals at various scales; then, these sub-signals are synthesized to obtain an approximate reconstruction of the original signal. In the inverse transformation process, the same wavelet basis functions and parameters as the forward transformation need to be used to ensure the precision and accuracy of the reconstruction.

[0155] 5) Synthesize the sub-signals at each scale: First, interpolate the sub-signals at each scale (using the cubic spline interpolation method: insert a cubic polynomial curve between two adjacent points so that the value and derivative of the curve at the two points are consistent with the original data. This method can better preserve the details of the original signal); then fuse with the low-frequency component to obtain an approximate reconstruction of the original signal (in the process of fusing with the low-frequency component, weighted processing can be performed according to the contribution of the sub-signals at different scales to the original signal, so that the fused signal is closer to the original signal, specifically including: performing wavelet transform on the sub-signals at each scale to obtain a set of wavelet coefficients; according to the statistical characteristics or features of the wavelet coefficients at each scale, the contribution of the wavelet coefficients to the original signal is determined by calculating the energy, variance, correlation and other indicators of the wavelet coefficients; according to the contribution of the wavelet coefficients at each scale, its weight is determined by normalization, fuzzy clustering, neural network and other methods; finally, the sub-signals at each scale are weighted superimposed with the corresponding weights to obtain a comprehensive signal)

[0156] 2. FRESH filtering technology

[0157] FRESH filtering technology and wavelet denoising technology are both signal processing technologies based on wavelet transform, which are used to remove noise and interference in signals.

[0158] FRESH filtering uses the Symmlet wavelet function; in order to better preserve the detail information of the image, FRESH filtering still uses soft thresholding processing.

[0159] Optionally, the signal filtering method is:

[0160] The signal to be processed is converted to the time-frequency domain using STFT (Short Time Fourier Transform), using a window length of 256 samples, an overlap rate of 75%, and a Hamming window. The length of the spectrum vector can be reduced to 129 by discarding frequency samples corresponding to negative frequencies. A fully convolutional network consisting of 16 convolutional layers is defined. The size of the convolution kernel of the first layer is 9×8, with 18 convolution kernels. The second to 13th convolutional layers are groups of 3 layers repeated 4 times, with convolution kernel widths of 5, 9, and 9, respectively, and the number of convolution kernels is 64. The convolution kernel width of the last convolution layer is 129 and there is only 1 convolution kernel. In this network, convolution is performed only in the frequency dimension, and for all layers except the first layer, the convolution kernel width along the time dimension is set to 1. The convolution layer is followed by a batch normalization layer and a ReLU activation function layer, and finally output through a regression layer.

[0161] A convolutional neural network is established to complete the mapping F between the STFT spectrum X(m, k) of the signal to be processed and the STFT spectrum S(m, k) of the clean signal to be processed, so as to achieve the purpose of denoising. The input prediction variable signal and the network target signal are the amplitude spectrum of the received noisy signal to be processed and the clean signal to be processed respectively. The goal of the network is to learn the mapping F so that the network outputs the amplitude spectrum of the noisy signal to be processed. The regression network uses the prediction variable input to minimize the mean square error between its output and the input target. Therefore, the loss function of the denoising network can be expressed as

[0162]

[0163] Among them, B represents the batch size, M and K represent the total number of STFT time frames and frequency points respectively. Finally, the denoised signal to be processed is converted back to the time domain signal using the amplitude spectrum output by the network and the phase of the received signal to be processed.

[0164] 3. Computer individual recognition method based on statistical distribution to distinguish different electromagnetic controlled terminals

[0165] The specific process includes:

[0166] 1) Statistical feature data collection: On the one hand, through various means such as active detection, passive monitoring or permission-based attribute information acquisition, various statistical feature data of multiple electromagnetic communication controlled terminals in the isolated target network are collected, including network traffic, system configuration, hardware information, operating system version, application version and other attribute information; on the other hand, through network monitoring tools or traffic analysis tools, various behavioral data of multiple electromagnetic communication controlled terminals in the isolated target network are collected, such as communication traffic, communication return time, communication volume, etc.

[0167] 2) Feature data preprocessing: By removing duplicate data, filtering invalid data, converting data formats and other processing methods, the collected statistical feature data such as behaviors and individuals are cleaned and standardized to facilitate subsequent analysis.

[0168] Data cleaning: solve the problems of missing values, abnormal values, and outliers; data integration: solve the problems of sample duplication, indicator construction, and highly similar attributes; data reduction: solve the problem of excessive data size; data transformation: convert data into data that is more convenient for analysis.

[0169] 3) Feature extraction: Extract features that can distinguish different electromagnetic communication controlled terminals from the collected statistical feature data. These features include behavioral statistical features such as communication behavior patterns and the size of communication electromagnetic waves; as well as individual statistical features such as the distribution of attribute information data and outliers;

[0170] 4) Construct a statistical learning model: Utilize the extracted features and corresponding classification labels (i.e., the device identification of the known electromagnetic communication controlled end) and establish a statistical model of the underlying data distribution for individual identification based on the generative adversarial network (GANs) in deep learning. Through learning and training, the statistical distribution describing the electromagnetic communication controlled end can be identified, thereby accurately identifying different known or unknown electromagnetic communication controlled end individuals.

[0171] 5) Classification and identification of electromagnetic communication controlled terminals: Use the trained statistical learning model to predict, classify and identify multiple electromagnetic communication controlled terminals in the current isolated network. Through the analysis and prediction of the current input samples, analyze the communication behavior patterns such as connection habits and data transmission modes, analyze the statistical characteristics such as the distribution of communication volume, communication frequency and transmission rate in different stages; analyze the individual statistical distribution characteristics such as application attributes, operating system attributes, and device attributes in different stages; and then comprehensively identify and determine the electromagnetic communication controlled terminals.

[0172] Optionally, the computer individual identification method based on statistical distribution can be used to distinguish different electromagnetic controlled terminals by the following steps:

[0173] 1) Collection of statistical characteristic data;

[0174] 2) Divide the statistical characteristic data to obtain source data and target data;

[0175] 3) Forward propagation: Use DCNN to obtain features from the source data, obtain instance graphs through the graph generation layer, and use Multireceptive Field Graph Convolutional Networks (MRF-GCN) to process the instance graphs to obtain the final feature parameters.

[0176] Instead of using a single convolution kernel on each layer, the Dynamic Convolutional Neural Networks (DCNN) dynamically aggregate multiple parallel convolution kernels based on attention. The dynamic convolution network introduces two parts of calculation: the attention model and the superposition of convolution kernels. Attention dynamically adjusts the weight of each convolution kernel according to the input, thereby generating adaptive dynamic convolution. The dynamic convolution network includes a feature extractor (F), a domain discriminator (D), and a label classifier (C). Get the predicted source data label from the label classifier Get the target data from the domain discriminator.

[0177] The dynamic convolution has K kernels, with attention weights {π k (x)} for aggregation. For different input images, the combination of kernels is also different. The attention weight is calculated as follows:

[0178]

[0179] Where g is the activation function, π k is the kth linear function Attention weight, aggregation weight and deviation is a function of the input and has the same attention. The dynamic convolution layer is constructed using batch normalization and activation function after the aggregate convolution, which can be used as a module to replace the simple convolution.

[0180] Undirected graph G = (V, X, A), where V = n represents a finite number of nodes. X is the feature matrix, A ∈ R n×n Represents the adjacency matrix of graph G. Graph data mainly consists of two key parts, feature matrix X and adjacency matrix A. The feature matrix is ​​obtained by feature extraction:

[0181] X = DCNN(X input )

[0182] Where X input is the mini-batch input matrix.

[0183] The adjacency matrix is ​​obtained using the Graph Generation Layer (GGL). First, the features are obtained from the source data using DCNN and input into a multilayer perceptron (MLP). Then the adjacency matrix is ​​obtained by multiplying the features of the MLP with its transpose. Finally, the top-k sorting mechanism is used to select the first k nearest neighbors of each node. It is expressed as follows:

[0184]

[0185] Where A is the constructed adjacency matrix, is the output of MLP, normlize(·) represents the normalization function, It is a sparse adjacency matrix, which reduces the computational burden.

[0186] The MRF-Conv layer contains several parallel graph convolutional layers with different receptive fields, which aggregate information from different numbers of neighbors. The MRF-Conv operation can be defined as:

[0187]

[0188] In the formula is a learnable parameter, [·] represents feature concatenation, H represents fusion features, and v represents the number of receptive fields of MRF-Conv.

[0189] 4) Back propagation: Calculate classification loss, domain alignment loss, and structural alignment loss, and use the Adam optimizer to update the dynamic convolutional network parameters;

[0190] Classification loss: In order to ensure that the label classifier achieves promising prediction results, the classification loss between the true label and the predicted label is estimated by the cross entropy loss, which is defined as:

[0191]

[0192] In the formula represents the prediction result of the label classifier, L(·,·) represents the cross entropy loss, represents mathematical expectation.

[0193] Domain alignment loss: Use the domain discriminator D to distinguish whether the extracted features are from the target domain or the source domain, and train the feature extractor to deceive the domain discriminator. Binary cross entropy loss is used as the domain alignment loss, expressed as:

[0194]

[0195] In the formula and They represent the feature extraction of the jth target sample and the ith source sample respectively. The value of D(·) takes 0 and 1, which can be used to distinguish which domain the sample comes from.

[0196] Structural alignment loss: In the process of aligning the feature structures of the source domain and the target domain, the GWD metric is used as the structural difference alignment loss. The GWD metric can measure the similarity of distribution regardless of whether the target sample and the original sample overlap.

[0197] 5) When the training time reaches the maximum value, the trained dynamic convolutional network is output;

[0198] 6) Use the trained dynamic convolutional network to test the target data and obtain the electromagnetic controlled terminal recognition results.

[0199] Figure 5 It is a working flow chart of the electromagnetic communication system disclosed in the embodiment of the present invention.

[0200] 4. Demodulation, decryption, decoding

[0201] 1)Unicode encoding and decoding:

[0202] Encoding process: First, you need to convert a character or string into a number. Unicode is a character set that assigns a unique number to each character. This number is the Unicode code point. Then, the Unicode code point is converted into binary data depending on the number of bytes required. For example, UTF-8 is a common Unicode encoding method that can use 1 to 4 bytes to represent a character.

[0203] Decoding process: After receiving the binary data, the receiver first determines the length of the byte.

[0204] Then, according to the length of the byte, the binary data is decoded into Unicode code points (the correspondence between Unicode code points and characters is defined by the Unicode encoding standard. Unicode is a character encoding standard that provides a unified encoding method for various languages ​​​​and symbols around the world.

[0205] In Unicode, each character has a unique Unicode code point, which is the position of the character in the Unicode encoding space. The Unicode encoding space is a continuous range of numbers that covers characters and symbols of all languages ​​in the world. Decoding needs to be done according to a specific encoding format. In Python, you can use the str.decode() method to decode binary data into a Unicode string; if you don't know the encoding format of the binary data, you can use the chardet module in Python to automatically detect it, and then use the detected encoding format to decode it).

[0206] Finally, convert the Unicode code point back to the original character or string (based on the Unicode code point decoded from a certain encoding (such as UTF-8). When a character or string is converted to a Unicode code point, it is actually mapped to a specific position in the Unicode encoding space. This position is the Unicode code point of the character or string; when the Unicode code point needs to be converted back to the original character or string, the reverse operation can be performed, that is, the corresponding character or string is found from the specific position in the Unicode encoding space. This process is based on the Unicode encoding standard and mapping rules, which ensures the unique correspondence between characters and Unicode code points.).

[0207] It should be noted that different Unicode encoding methods (such as UTF-8, UTF-16, etc.) may differ in the encoding and decoding process. But the basic steps are similar: convert the character or string to a number, then convert the number to binary data, and finally decode the binary data back to a character or string.

[0208] 2) AES256 algorithm decryption

[0209] The AES256 algorithm is a symmetric encryption algorithm that uses a 256-bit key for encryption and decryption operations. The encryption and decryption process of the AES256 algorithm is as follows:

[0210] Encryption process:

[0211] a) Choose a 256-bit key.

[0212] b) Divide the plaintext into 128-bit data blocks. If the last data block is less than 128 bits, add zeros to the end of the data block.

[0213] c) Each data block is encrypted using a key. The encryption process includes multiple steps such as byte replacement, row shift, column obfuscation, and round key addition.

[0214] Byte replacement: Replace each byte in the plaintext with a fixed replacement value.

[0215] Row shift: Shift each row in the plaintext in a specific way.

[0216] Column obfuscation: Obfuscate each column in the plaintext so that the value of each byte is affected by other bytes.

[0217] Round key addition: XOR each byte with a round key to increase the strength of the encryption.

[0218] d) Connect the encrypted data blocks to form ciphertext.

[0219] Decryption process:

[0220] a) Decrypt the ciphertext using the same 256-bit key.

[0221] b) The decryption process is the opposite of the encryption process, including byte replacement, row shift, column confusion and round key reduction steps.

[0222] c) Inverse byte replacement: Replace each byte in the ciphertext with its corresponding inverse replacement value.

[0223] d) Reverse shift: Shift each row in the ciphertext in the opposite way to that used during encryption.

[0224] e) Reverse column obfuscation: Perform a reverse obfuscation operation on each column in the ciphertext so that the value of each byte is restored to its original state.

[0225] f) Round key subtraction: XOR each byte with a round key to recover the original data.

[0226] g) Connect the decrypted data blocks to form plaintext.

[0227] 3) Frequency Shift Keying (FSK) Modulation and Demodulation

[0228] The FSK modulation and demodulation process includes the following steps:

[0229] a) Modulation: At the controlled end of electromagnetic communication, the original data stream is mapped to two or more frequencies of FSK. For example, if two frequencies f1 and f2 are used to represent data, 0 can be mapped to f1 and 1 can be mapped to f2.

[0230] b) Input the raw data bit stream into the FSK modulator.

[0231] According to the value of the data bit, the corresponding frequency is selected as the carrier. For example, if the data bit is 0, the frequency f1 can be selected as the carrier; if the data bit is 1, the frequency f2 can be selected as the carrier.

[0232] The selected carrier signal is generated by an oscillator and amplified by a power amplifier before output.

[0233] c) Transmission: The modulated signal is sent to the receiving end through the channel.

[0234] d) Demodulation: At the electromagnetic communication master end, the received signal is sent to the frequency detector after front-end amplification, filtering and other processing. The frequency detector detects the frequency components of the received signal and determines which frequency corresponds to which data symbol based on the preset frequency standard. For example, if the frequency detector detects that the frequency of the received signal is f1, it can be determined that the signal represents data symbol 0; if the frequency is detected to be f2, it can be determined that the signal represents data symbol 1.

[0235] e) After the received FSK signal is processed by front-end amplification, filtering, etc., it is input into the FSK demodulator.

[0236] Front-end amplifier includes:

[0237] Signal reception: The front-end amplifier first receives the signal from the antenna. These signals may be very weak, so they need to be amplified for subsequent processing.

[0238] Signal amplification: The front-end amplifier uses an amplifier circuit to amplify the received signal. The amplification factor can be adjusted as needed to adapt to different signal strengths.

[0239] Noise suppression: While amplifying the signal, the front-end amplifier also suppresses noise and interference. This helps improve the quality of the signal and reduces the burden on subsequent processing circuits.

[0240] The filtering process includes:

[0241] Signal input: The filter receives signals from the front-end amplification. These signals may contain multiple frequency components, some of which may be noise or interference.

[0242] Frequency selection: The filter selects the frequency components it allows to pass based on its frequency characteristics. This is usually achieved by using a network of components such as resistors, capacitors, and inductors.

[0243] Noise suppression: Filters can suppress unwanted frequency components such as noise and interference. This helps improve the quality of the signal and reduce misjudgment in subsequent demodulation circuits.

[0244] The frequency component of the received signal is detected using a frequency detector.

[0245] According to the preset frequency standard, it is determined which data symbol the frequency of the received signal belongs to. For example, if the frequency detector detects that the frequency of the received signal is f1, it can be determined that the signal represents data symbol 0; if the frequency is detected to be f2, it can be determined that the signal represents data symbol 1.

[0246] The determined data symbol is output to complete the demodulation of the FSK signal.

[0247] Embodiment 3

[0248] See also Figure 6 , Figure 6 : is a schematic diagram of a VGA-oriented electromagnetic communication device disclosed in an embodiment of the present invention. Figure 6 The VGA-oriented electromagnetic communication device described is used to establish a one-way data return channel in an isolation network, and can also be used in conjunction with other means, such as sound waves, to establish a two-way data transmission channel, which is not limited in the embodiments of the present invention. Figure 6 As shown, the VGA-oriented electromagnetic communication device may include the following operations:

[0249] S301, an electromagnetic communication controlled terminal selection module, used to perform time-sharing processing on N electromagnetic communication controlled terminals in the same period and network, and select an electromagnetic communication controlled terminal for communication, including:

[0250] S11, collecting information from the electromagnetic communication controlled terminals in the same period and network to obtain characteristic information of the electromagnetic communication controlled terminals;

[0251] The characteristic information includes attribute information and behavior information;

[0252] The attribute information includes network traffic information, system configuration information, hardware information, operating system version information, and application version information;

[0253] The behavior information includes communication flow information and communication return time information;

[0254] S12, preprocessing the feature information to obtain preprocessed feature information, including:

[0255] S121, performing data cleaning on the feature information to remove missing values ​​and outliers to obtain first feature information;

[0256] S122, constructing data indicators for the first characteristic information to obtain second characteristic information;

[0257] S123, performing data standardization processing on the second characteristic information to obtain third characteristic information;

[0258] S124, processing the third feature information to obtain pre-processed feature information;

[0259] S13, using the preprocessed feature information, training a preset data underlying distribution statistical model to obtain an optimized data underlying distribution statistical model;

[0260] S14, using the optimized data underlying distribution statistical model, predicting N electromagnetic communication controlled terminals in the same network at the same time period in the isolated network to be processed, and selecting an electromagnetic communication controlled terminal for communication;

[0261] The electromagnetic communication controlled terminal for communication includes a VGA interface;

[0262] S302, a data sending module, used to send data to the electromagnetic communication master terminal by using the VGA interface of the electromagnetic communication controlled terminal for communication;

[0263] S303, a data restoration module, used to restore the received data using the electromagnetic communication main control terminal to implement electromagnetic communication, including:

[0264] S31, using the electromagnetic communication main control terminal, performing denoising processing on the received data to obtain denoised data information;

[0265] S32, filtering the denoised data information to obtain filtered data information;

[0266] The signal filtering method is:

[0267] The signal to be processed is converted into the time-frequency domain using STFT, using a window length of 256 samples, an overlap of 75% and a Hamming window;

[0268] Reduce the length of the spectrum vector to 129 by discarding frequency samples corresponding to negative frequencies;

[0269] Define a fully convolutional network consisting of 16 convolutional layers. The first layer has a convolution kernel size of 9×8 and 18 convolution kernels. The second to 13th convolutional layers are a group of 3 layers repeated 4 times, with convolution kernel widths of 5, 9, and 9 respectively, and the number of convolution kernels is 64. The convolution kernel width of the last convolution layer is 129 and there is only 1 convolution kernel. In this network, convolution is performed only in the frequency dimension, and for all layers except the first layer, the convolution kernel width along the time dimension is set to 1. After the convolution layer, it passes through a batch normalization layer and a ReLU activation function layer, and finally outputs through a regression layer;

[0270] Establish a convolutional neural network to complete the mapping F between the STFT spectrum X(m, k) of the signal to be processed and the STFT spectrum S(m, k) of the clean signal to be processed, so as to achieve the purpose of denoising;

[0271] The loss function of the denoising network is expressed as:

[0272]

[0273] Among them, B represents the batch size, M and K represent the total number of STFT time frames and frequency points respectively; finally, the denoised signal to be processed is converted back to the time domain signal using the amplitude spectrum output by the network and the phase of the received signal to be processed;

[0274] S33, demodulating the filtered data information to obtain demodulated data information;

[0275] S34, decrypting the demodulated data information to obtain decrypted data information;

[0276] S35, decoding the decrypted data information to obtain decoded data information, and realizing electromagnetic communication.

[0277] Embodiment 4

[0278] See also Figure 7 , Figure 7 : is a schematic diagram of another VGA-oriented electromagnetic communication device disclosed in an embodiment of the present invention. Figure 7 The VGA-oriented electromagnetic communication device described is used to establish a one-way data return channel in an isolation network, and can also be used in conjunction with other means, such as sound waves, to establish a two-way data transmission channel, which is not limited in the embodiments of the present invention. Figure 7 As shown, the VGA-oriented electromagnetic communication device may include the following operations:

[0279] A memory 401 storing executable program codes;

[0280] a processor 402 coupled to the memory 401;

[0281] The processor 402 calls the executable program code stored in the memory 401 to execute the steps of the VGA-oriented electromagnetic communication method described in the first and second embodiments.

[0282] Embodiment 5

[0283] An embodiment of the present invention discloses a computer-readable storage medium storing a computer program for electronic data exchange, wherein the computer program enables a computer to execute the steps of the VGA-oriented electromagnetic communication method described in the first and second embodiments.

[0284] The device embodiments described above are only illustrative, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, i.e., they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative work.

[0285] Through the specific description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the above technical solution can be essentially or partly contributed to the prior art in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0286] Finally, it should be noted that the VGA-oriented electromagnetic communication method and device disclosed in the embodiments of the present invention disclose only the preferred embodiments of the present invention, which are only used to illustrate the technical solution of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A VGA-oriented electromagnetic communication method, characterized in that: The method comprises: S1, performing time-sharing processing on N electromagnetic communication controlled terminals in the same period and network, and selecting the electromagnetic communication controlled terminal for communication, including: S11, collecting information from the electromagnetic communication controlled terminals in the same period and network to obtain characteristic information of the electromagnetic communication controlled terminals; The characteristic information includes attribute information and behavior information; The attribute information includes network traffic information, system configuration information, hardware information, operating system version information, and application version information; The behavior information includes communication flow information and communication return time information; S12, preprocessing the feature information to obtain preprocessed feature information, including: S121, performing data cleaning on the feature information to remove missing values ​​and outliers to obtain first feature information; S122, constructing data indicators for the first characteristic information to obtain second characteristic information; S123, performing data standardization processing on the second characteristic information to obtain third characteristic information; S124, processing the third feature information to obtain pre-processed feature information; S13, using the preprocessed feature information, training a preset data underlying distribution statistical model to obtain an optimized data underlying distribution statistical model; S14, using the optimized data underlying distribution statistical model, predicting N electromagnetic communication controlled terminals in the same network at the same time period in the isolated network to be processed, and selecting an electromagnetic communication controlled terminal for communication; The electromagnetic communication controlled terminal for communication includes a VGA interface; S2, using the VGA interface of the electromagnetic communication controlled terminal for communication, sending data to the electromagnetic communication master terminal; S3, using the electromagnetic communication main control terminal to restore the received data to achieve electromagnetic communication, including: S31, using the electromagnetic communication main control terminal, performing denoising processing on the received data to obtain denoised data information; S32, filtering the denoised data information to obtain filtered data information; The signal filtering method is: The signal to be processed is converted into the time-frequency domain using STFT, using a window length of 256 samples, an overlap of 75% and a Hamming window; Reduce the length of the spectrum vector to 129 by discarding frequency samples corresponding to negative frequencies; Define a fully convolutional network consisting of 16 convolutional layers. The first layer has a convolution kernel size of 9×8 and 18 convolution kernels. The second to 13th convolutional layers are a group of 3 layers repeated 4 times, with convolution kernel widths of 5, 9, and 9 respectively, and the number of convolution kernels is 64. The convolution kernel width of the last convolution layer is 129 and there is only 1 convolution kernel. In this network, convolution is performed only in the frequency dimension, and for all layers except the first layer, the convolution kernel width along the time dimension is set to 1. After the convolution layer, it passes through a batch normalization layer and a ReLU activation function layer, and finally outputs through a regression layer; Establish a convolutional neural network to complete the mapping F between the STFT spectrum X(m, k) of the signal to be processed and the STFT spectrum S(m, k) of the clean signal to be processed, so as to achieve the purpose of denoising; The loss function of the denoising network is expressed as: Among them, B represents the batch size, M and K represent the total number of STFT time frames and frequency points respectively; finally, the denoised signal to be processed is converted back to the time domain signal using the amplitude spectrum output by the network and the phase of the received signal to be processed; S33, demodulating the filtered data information to obtain demodulated data information; S34, decrypting the demodulated data information to obtain decrypted data information; S35, decoding the decrypted data information to obtain decoded data information, thereby realizing electromagnetic communication.

2. A VGA-oriented electromagnetic communication device, characterized in that: The device comprises: The electromagnetic communication controlled terminal selection module is used to perform time-sharing processing on N electromagnetic communication controlled terminals in the same period and network, and select the electromagnetic communication controlled terminal for communication, including: S11, collecting information from the electromagnetic communication controlled terminals in the same period and network to obtain characteristic information of the electromagnetic communication controlled terminals; The characteristic information includes attribute information and behavior information; The attribute information includes network traffic information, system configuration information, hardware information, operating system version information, and application version information; The behavior information includes communication flow information and communication return time information; S12, preprocessing the feature information to obtain preprocessed feature information, including: S121, performing data cleaning on the feature information to remove missing values ​​and outliers to obtain first feature information; S122, constructing data indicators for the first characteristic information to obtain second characteristic information; S123, performing data standardization processing on the second characteristic information to obtain third characteristic information; S124, processing the third feature information to obtain pre-processed feature information; S13, using the preprocessed feature information, training a preset data underlying distribution statistical model to obtain an optimized data underlying distribution statistical model; S14, using the optimized data underlying distribution statistical model, predicting N electromagnetic communication controlled terminals in the same network at the same time period in the isolated network to be processed, and selecting an electromagnetic communication controlled terminal for communication; The electromagnetic communication controlled terminal for communication includes a VGA interface; A data sending module, used to send data to the electromagnetic communication master terminal by using the VGA interface of the electromagnetic communication controlled terminal for communication; The data restoration module is used to restore the received data using the electromagnetic communication main control terminal to realize electromagnetic communication, including: S31, using the electromagnetic communication main control terminal, performing denoising processing on the received data to obtain denoised data information; S32, filtering the denoised data information to obtain filtered data information; The signal filtering method is: The signal to be processed is converted into the time-frequency domain using STFT, using a window length of 256 samples, an overlap of 75% and a Hamming window; Reduce the length of the spectrum vector to 129 by discarding frequency samples corresponding to negative frequencies; Define a fully convolutional network consisting of 16 convolutional layers. The first layer has a convolution kernel size of 9×8 and 18 convolution kernels. The second to 13th convolutional layers are a group of 3 layers repeated 4 times, with convolution kernel widths of 5, 9, and 9 respectively, and the number of convolution kernels is 64. The convolution kernel width of the last convolution layer is 129 and there is only 1 convolution kernel. In this network, convolution is performed only in the frequency dimension, and for all layers except the first layer, the convolution kernel width along the time dimension is set to 1. After the convolution layer, it passes through a batch normalization layer and a ReLU activation function layer, and finally outputs through a regression layer; Establish a convolutional neural network to complete the mapping F between the STFT spectrum X(m, k) of the signal to be processed and the STFT spectrum S(m, k) of the clean signal to be processed, so as to achieve the purpose of denoising; The loss function of the denoising network is expressed as: Among them, B represents the batch size, M and K represent the total number of STFT time frames and frequency points respectively; finally, the denoised signal to be processed is converted back to the time domain signal using the amplitude spectrum output by the network and the phase of the received signal to be processed; S33, demodulating the filtered data information to obtain demodulated data information; S34, decrypting the demodulated data information to obtain decrypted data information; S35, decoding the decrypted data information to obtain decoded data information, thereby realizing electromagnetic communication.

3. A VGA-oriented electromagnetic communication device, characterized in that: The device comprises: A memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the VGA-oriented electromagnetic communication method as claimed in claim 1.

4. A computer storable medium, characterized in that: The computer storable medium stores computer instructions, and when the computer instructions are called, they are used to execute the VGA-oriented electromagnetic communication method as claimed in claim 1.

Images