Wireless communication secure transmission method, device and equipment based on deep learning

Abstract

The invention discloses a wireless communication secure transmission method, device and equipment based on deep learning. Comprising the following steps: a transmitter performs one-hot coding on a to-be-sent message of each channel to obtain a plurality of one-hot coding matrixes; inputting the one-hot coding matrix and the channel characteristic sequence into a corresponding automatic encoder to obtain a plurality of modulation coding signals; performing zero-forcing pre-coding on the plurality of modulation coding signals to generate a plurality of pre-coding signals; performing power constraint on the plurality of pre-coded signals to obtain a transmitting signal, and transmitting the transmitting signal to a receiver; and the receiver obtains a receiving signal transmitted by the target channel, and decodes the receiving signal based on an automatic decoder to obtain decoded data. According to the invention, the reliability of information transmission is ensured by using the automatic codec, the design and implementation of the communication system are simplified, and the applicability and efficiency of the communication system in diversified environments are improved.

Description

Technical Field

[0001] The present application relates to the field of wireless communication technology, and in particular to a method, device and equipment for secure transmission of wireless communication based on deep learning. Background Art

[0002] Physical layer security is the cornerstone of communication system security. With the development of communication technology, especially the advent of the 5G and 6G era, wireless communication networks have become more complex and open, which increases the risk of information being intercepted and cracked. Physical layer security ensures the security of information by introducing protection measures during signal transmission, such as signal design, coding, and modulation. This approach does not rely on complex key management and distribution, but instead uses the physical characteristics of the communication channel to achieve secure communication, which is critical to preventing unauthorized access and protecting sensitive data. Physical layer security can provide a way of communication that remains secure even in the face of attackers with powerful computing power.

[0003] At present, traditional methods usually need to design encoding and decoding schemes based on complex mathematical models and algorithms, which requires deep expertise and experience. In addition, in the face of changing channel conditions and external interference, traditional methods usually require manual redesign and adjustment of parameters, which is not only time-consuming but also prone to errors. Summary of the invention

[0004] The embodiments of the present application provide a method, device and apparatus for secure transmission of wireless communication based on deep learning, so as to at least solve the technical problem in the related art that it is difficult to automatically encode and decode the communication system.

[0005] According to one aspect of an embodiment of the present application, a wireless communication security transmission method based on deep learning is provided, which is used for a transmitter, including:

[0006] Perform one-hot encoding on the messages to be sent on each channel to obtain multiple one-hot encoding matrices;

[0007] Inputting the one-hot coding matrix and the channel feature sequence into a corresponding automatic encoder to obtain a plurality of modulated coded signals;

[0008] Performing zero-forcing precoding on the multiple modulated coded signals to generate multiple precoded signals;

[0009] Power constraints are performed on the multiple precoded signals to obtain a transmit signal, and the transmit signal is transmitted to a receiver.

[0010] In one embodiment, before the one-hot encoding matrix and the channel feature sequence are input into the corresponding autoencoder, the method further includes:

[0011] The autoencoder is trained based on a fully connected neural network to obtain the trained autoencoder.

[0012] In one embodiment, training the autoencoder based on a fully connected neural network includes:

[0013] Randomly generate a preset number of channel matrices:

[0014] Initialize the autoencoder network, including the transmitter and multiple receiver subnetworks;

[0015] Randomly generate a confidential message and generate an encoded transmission signal based on the transmitter network function;

[0016] Transmitting the transmission signal to a receiver through a channel layer, and adding Gaussian white noise at the receiver;

[0017] Using an automatic decoder to decode the estimated confidential message based on the received signal and the channel characteristic sequence;

[0018] The transmitter and multiple receiver sub-networks are trained simultaneously until the model converges to obtain a trained autoencoder.

[0019] In one embodiment, it further includes:

[0020] Based on average cross entropy as the loss function for each receiver;

[0021] Based on the sum of the loss functions of multiple receivers, the total loss function is obtained.

[0022] In one implementation, power constraints are performed on the multiple precoded signals to obtain a transmission signal, including:

[0023] Introducing power splitting factors to distribute the total power to multiple receivers;

[0024] Calculating a power limit factor for each precoded signal based on the divided power, and scaling each precoded signal based on the power limit factor;

[0025] Based on the sum of the scaled precoded signals, a transmit signal is obtained.

[0026] According to another aspect of an embodiment of the present application, a wireless communication security transmission method based on deep learning is provided, which is used in a receiver, including:

[0027] Obtain the received signal transmitted from the target channel;

[0028] The received signal is decoded based on an automatic decoder to obtain decoded data.

[0029] In one implementation, after obtaining the decoded data, the method further includes:

[0030] Input the decoded data into a preset activation function to obtain an output probability matrix;

[0031] Based on the probability matrix, the codeword with the highest probability is selected as the final output.

[0032] According to another aspect of an embodiment of the present application, a wireless communication secure transmission method based on deep learning is provided, including:

[0033] The transmitter performs one-hot encoding on the message to be sent on each channel to obtain multiple one-hot encoding matrices; inputs the one-hot encoding matrix and the channel characteristic sequence into the corresponding autoencoder to obtain multiple modulated coded signals; performs zero-forcing precoding on the multiple modulated coded signals to generate multiple precoded signals; performs power constraints on the multiple precoded signals to obtain transmission signals, and transmits the transmission signals to the receiver;

[0034] The receiver acquires a received signal transmitted from a target channel, and decodes the received signal based on an automatic decoder to obtain decoded data.

[0035] According to another aspect of an embodiment of the present application, a wireless communication secure transmission device based on deep learning is provided, comprising:

[0036] The transmitter is used to perform one-hot encoding on the message to be sent on each channel to obtain multiple one-hot encoding matrices; input the one-hot encoding matrix and the channel characteristic sequence into the corresponding automatic encoder to obtain multiple modulated coded signals; perform zero-forcing precoding on the multiple modulated coded signals to generate multiple precoded signals; perform power constraints on the multiple precoded signals to obtain transmission signals, and transmit the transmission signals to the receiver;

[0037] The receiver is used to obtain a received signal transmitted from a target channel, and decode the received signal based on an automatic decoder to obtain decoded data.

[0038] According to another aspect of an embodiment of the present application, there is also provided an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to execute the above-mentioned deep learning-based wireless communication security transmission method through the computer program.

[0039] The technical solution provided by the embodiments of the present application may have the following beneficial effects:

[0040] The deep learning-based autoencoder of the present application can automatically learn the encoding and decoding process of the signal, reducing the reliance on complex mathematical models in traditional communication theory. This makes the system design more automated and reduces the reliance on expert knowledge. In addition, the adaptability of the autoencoder enables it to maintain optimal performance by adjusting its parameters in the face of changing channel conditions and external interference, which usually requires manual redesign and adjustment of parameters in traditional methods, which is not only time-consuming but also error-prone. This improvement in computing efficiency is particularly important for the deployment of large-scale communication networks because it allows more devices to communicate with lower energy consumption and cost. Therefore, the autoencoder not only simplifies the design and implementation of the communication system, but also improves its applicability and efficiency in diverse environments. The method uses an automatic codec to ensure the reliability of information transmission, and uses zero-forcing precoding technology to achieve secure communication between multiple users, ensuring the secure transmission of information between nodes in the communication system. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0042] Figure 1 A wireless communication security transmission method for a transmitter according to an embodiment of the present application;

[0043] Figure 2 A wireless communication security transmission method for a receiver according to an embodiment of the present application;

[0044] Figure 3 A wireless communication security transmission method according to an embodiment of the present application;

[0045] Figure 4 is a schematic diagram of a system model of a dual-user MIMO broadcast channel according to an embodiment of the present application;

[0046] Figure 5 is a schematic diagram of a deep self-encoder architecture of a dual-user MIMO broadcast channel according to an embodiment of the present application;

[0047] Figure 6 It is a schematic diagram of the structure of an optional electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0048] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0049] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0050] In modern communication systems, there is a need to securely transmit information in the presence of multiple receivers, while ensuring that only the intended receiver can decrypt the information, while other receivers (including legitimate but non-intended receivers and potential eavesdroppers) cannot obtain the information. In this scenario, the sender needs to send independent messages to each intended receiver and ensure that these messages are confidential to other receivers. This requires the sender to take specific security measures during the encoding and transmission process to prevent information leakage.

[0051] Multi-antenna systems bring high complexity to signal processing, and many solutions rely on specific channel models and assumptions, which limits their applicability and flexibility in the face of changes and uncertainties in actual communication environments. By jointly optimizing modulation and precoding strategies, deep autoencoders can enhance the confidentiality of information while maintaining the reliability of information transmission, and achieve near-optimal confidentiality capacity in low signal-to-noise ratio areas. In addition, the end-to-end learning capability of the autoencoder allows the system to learn transmission and reception strategies directly from the data without the need to optimize each part of the communication link separately, thereby improving the adaptability and overall performance of the system. Therefore, the use of deep autoencoders improves the reliability of multi-user communication scenarios, while reducing computational complexity and improving the overall performance of the system. In addition, in order to avoid information leakage between confidential messages and unintended users, the present application adds zero-forcing precoding in the transmitter to ensure the confidentiality of confidential message transmission.

[0052] Currently, simultaneous and confidential communication is generally achieved through secret dirty paper coding (S-DPC). However, S-DPC is too complex for practical applications. In addition, there are also a variety of deep learning-based methods, but most of these methods use deep learning combined with time division multiple access (TDMA) to transmit information to different users. In this mode, one user can intercept the communication content of another user, and the confidentiality of information transmission cannot be guaranteed.

[0053] Based on this, the embodiment of the present application provides a wireless communication security transmission method based on deep learning. Figure 1-5 The wireless communication security transmission method based on deep learning in the embodiment of the present application is introduced in detail. Figure 1 As shown, a wireless communication security transmission method based on deep learning, used in a transmitter, comprises the following steps:

[0054] S101 performs one-hot encoding on the messages to be sent on each channel to obtain multiple one-hot encoding matrices.

[0055] In one embodiment, when there are multiple receivers, the transmitter needs to send different confidential messages to the corresponding receivers through multiple channels. First, the message to be sent on each channel needs to be one-hot encoded to obtain a one-hot encoding matrix for each channel.

[0056] One-Hot Encoding is a method for converting categorical data into numerical form, so that machine learning algorithms can better process this data. In one-hot encoding, each category value is converted into a binary vector, with all positions except one representing the category being 1 and all other positions being 0. This encoding method is often used to process categorical variables, especially in the fields of natural language processing and computer vision.

[0057] For example, Figure 5 , which is a schematic diagram of a deep autoencoder architecture for a dual-user MIMO broadcast channel. The transmission system includes a bit input, a transmitter, a channel, a receiver, and a bit output.

[0058] The transmitter sends R1 and R2 bit messages W1 and W2 to receiver 1 and receiver 2 respectively. Message W i It is encoded into a binary one-hot encoding matrix P before entering the autoencoder i . Perform one-hot encoding on message W1 to obtain one-hot encoding matrix P1, and perform one-hot encoding on message W2 to obtain one-hot encoding matrix P2.

[0059] It is understandable that the present application does not limit the number of receivers and messages, and there may be multiple messages in actual applications. Each confidential message is uniquely encoded.

[0060] S102 inputs the one-hot coding matrix and the channel feature sequence into the corresponding automatic encoder to obtain multiple modulated coding signals.

[0061] In the embodiment of the present application, in addition to the bits used for message transmission, the input also adds channel characteristic sequences v1 and v2 so that the network can perform encoding and decoding. The definition of the channel characteristic sequence is as follows:

[0062]

[0063]

[0064] Among them, vec(·) represents the vectorization of the matrix, and are the null spaces of H1 and H2 respectively, and the sizes of v1 and v2 are n1(n t -n2) and n2(n t -n1). Equipped with n t ,n1 and n2 antennas, where n t ≥n1+n2.

[0065] In the dual-user system of the embodiment of the present application, two autoencoders are constructed respectively through a fully connected neural network, and the modulation and encoding of the confidential message are realized through the channel feature sequence. The one-hot encoding matrix and the channel feature sequence are input into the corresponding autoencoder to obtain multiple modulated coded signals.

[0066] It can be understood that the transmitter can also have multiple autoencoders. For example, if there are three receivers, three autoencoders can be constructed, and the one-hot coding matrix of each channel and the channel feature sequence can be input into the corresponding autoencoder to obtain three modulated coded signals.

[0067] like Figure 5 The dual-user system shown in the figure obtains two modulation coded signals: and Can be designed as s i and v i Function:

[0068] u i =E i (s i ,v i )

[0069] Among them, E i represents the function of the i-th autoencoder, which consists of a G-layer neural network. Therefore, u i The calculation process can be expressed as:

[0070]

[0071] in, Indicates E i The calculation of the g-th layer of the neural network, and They represent activation functions, E i Weights and biases for layer g.

[0072] S103 performs zero-forcing precoding on multiple modulated coded signals to generate multiple precoded signals.

[0073] like Figure 5 As shown in the two-user system, in order to ensure that the message received by the intended recipient is confidential and the unintended recipient cannot decode the confidential information, the transmitter modulates the coded signal u i Null space sent to unintended receivers. Precoded signal and It can be expressed as:

[0074]

[0075]

[0076] It can be understood that when there are multiple modulation and coding signals, zero-forcing precoding is performed on the multiple modulation and coding signals to generate multiple precoded signals. For example, if there are 5 modulation and coding signals, zero-forcing precoding is performed on the 5 modulation and coding signals to obtain 5 precoded signals.

[0077] Among them, Zero Forcing (ZF) is a linear precoding technology used in Multiple Input Multiple Output (MIMO) systems. Its core idea is to separate the transmitted data streams through beamforming by the base station, so that a unit response is generated in the desired direction and the response in the undesired direction is zero, thereby eliminating interference between user data streams.

[0078] S104 performs power constraints on the multiple precoded signals to obtain a transmission signal, and transmits the transmission signal to a receiver.

[0079] In one implementation, power constraints are imposed on a plurality of precoded signals to obtain a transmission signal, including: introducing a power division factor to distribute total power to a plurality of receivers.

[0080] Specifically, a power split factor γ∈[0,1] is introduced, and γP t and (1-γ)P t The confidential message of receiver 1 and the confidential message of receiver 2 are independently assigned to each other. t Indicates total power.

[0081] Further, based on the divided power, a power limit factor of each precoded signal is calculated, and based on the power limit factor, each precoded signal is scaled respectively.

[0082] Specifically, the power limit factor of each precoding signal is calculated, including:

[0083]

[0084]

[0085] The precoded signals s1 and s2 are scaled by power limiting factors β1 and β2 to achieve desired power control.

[0086] Based on the sum of the scaled precoded signals, the transmitted signal X is obtained.

[0087] The final output of the transmitter is:

[0088]

[0089] It can be seen that the average power of signals β1s1 and β2s2 is limited to γP t and (1-γ)P t Within, and the output X satisfies the total average power P t When there are multiple precoded signals, they are also scaled according to the above method to obtain the transmitted signal.

[0090] In one embodiment, before inputting the one-hot encoding matrix and the channel feature sequence into the corresponding autoencoder, the method further includes: training the autoencoder based on a fully connected neural network to obtain a trained autoencoder.

[0091] Specifically, a preset number of groups of channel matrices are randomly generated; an autoencoder network is initialized, including a transmitter and a plurality of receiver subnetworks; a confidential message is randomly generated, and an encoded transmission signal is generated based on a transmitter network function; the transmission signal is transmitted to a receiver through a channel layer, and Gaussian white noise is added at the receiver; an automatic decoder is used to decode the estimated confidential message based on the received signal and a channel feature sequence; and the transmitter and the plurality of receiver subnetworks are simultaneously trained until the model converges to obtain a trained autoencoder.

[0092] The present application randomly generates M pairs of channel matrices. During the training process, the channel layer is a non-trainable layer between the transmitter and the two receivers. The channel matrices serve as weights for the channel layers of receiver 1 and receiver 2, respectively. The transmitter function is responsible for encoding the confidential message into the transmitted signal X. After the signal passes through the channel layer, Gaussian white noise is added at the two receivers, respectively. The three subnetworks of the transmitter and receiver are trained simultaneously. Finally, when the SER indicators of both receivers do not decrease, the model converges. In an exemplary scenario, the autoencoder training algorithm for the dual-user system is as follows:

[0093] Randomly generate M pairs of channel matrices.

[0094] Initialize the autoencoder network.

[0095] for i=1:epochs do; Outer loop (epochs): Perform multiple iterations (epochs), in each iteration, the network will be trained through all channel matrices;

[0096] for i=1:M do; Inner loop, the following steps are executed M times.

[0097] Randomly generate secret messages W1 and W2 with block length n;

[0098] P1←One-hot encoder(W1); One-hot encode the confidential message W1 to obtain the matrix P1;

[0099] P2←One-hot encoder(W2); One-hot encode the confidential message W2 to obtain the matrix P2;

[0100] The transmitter function is used to generate the transmission signal based on the matrices P1, P2 and the channel characteristic sequences v1, v2; at the two receivers, the transmission signal X is passed through the corresponding channel matrix, and then Gaussian white noise z1 and z2 are added to simulate the channel transmission and noise interference in actual communication to obtain the received signal;

[0101] Using the automatic decoder and According to the received signal and the channel characteristic sequence, the estimated confidential message is decoded and

[0102] Calculate the overall loss value

[0103] Use Adam to update the weights of DAE-BCCM.

[0104] end for;

[0105] end for.

[0106] In one embodiment, the method further includes: constructing a loss function in the training process, including: using the average cross entropy as the loss function of each receiver; and obtaining a total loss function based on the sum of the loss functions of multiple receivers.

[0107] Each receiver has an estimate of the transmitted bits. Therefore, the overall loss function is in and are the loss functions at receiver 1 and receiver 2 respectively. In order to enable both receivers to correctly recover the confidential message transmitted by the transmitter, the average cross entropy (ACE) is used as the loss function to train the network:

[0108]

[0109] Among them, i∈{1,2} represents the i-th user, Represents confidential message W i The one-hot encoding matrix The jth row and the wth i For example, when w i =1 is expected message, and represents the probability matrix estimated at receiver i The jth row and the wth i The element corresponding to the column.

[0110] The loss function treats the receiver's output message codeword as a multi-classification problem, and calculates the difference between the predicted probability distribution of the codeword and the probability distribution of the actual codeword. The average sum of losses of two receivers when receiving n codewords is calculated.

[0111] like Figure 2 As shown, a wireless communication security transmission method based on deep learning, used in a receiver, comprises the following steps:

[0112] S201 obtains a received signal transmitted from a target channel.

[0113] In an optional implementation, there are two receivers. After the transmission signal X is transmitted through two channels respectively, the received signals at the two receivers are Y1 and Y2. Y1 and Y2 can be expressed as:

[0114]

[0115]

[0116] It can be seen that the received signals Y1 and Y2 do not contain any unexpected message components, and the transmitter design meets the requirements of confidential transmission.

[0117] The method provided in the embodiment of the present application does not limit the number of receivers, which can be set according to actual conditions.

[0118] S202 decodes the received signal based on the automatic decoder to obtain decoded data.

[0119] Each receiver contains an automatic decoder. and Responsible for decoding the received signals Y1 and Y2. and It is constructed by a Q-layer fully connected neural network, which takes the received signals Y1 and Y2 and the channel feature sequences v1 and v2 as input. The decoded signal d i It can be expressed as:

[0120]

[0121] in, express The calculation of the qth layer of the neural network, and They represent activation functions, Weights and biases for the qth layer.

[0122] In one embodiment, after obtaining the decoded data, the method further includes: inputting the decoded data into a preset activation function to obtain an output probability matrix; and based on the probability matrix, selecting the codeword with the highest probability as the final output.

[0123] exist and The last layer uses a Softmax activation function to map the output to a probability matrix, namely and The number of rows n represents the length of the message codeword, and the sum of the elements in each row is 1. i The column indicates that the codeword of the message in this row is w i Finally, receiver i selects the codeword with the highest probability as the final output based on the argmax function

[0124] like Figure 3 As shown, a wireless communication security transmission method based on deep learning is provided, including:

[0125] The S301 transmitter performs one-hot encoding on the messages to be sent on each channel to obtain multiple one-hot encoding matrices; inputs the one-hot encoding matrices and the channel characteristic sequence into the corresponding automatic encoder to obtain multiple modulated coding signals; performs zero-forcing precoding on the multiple modulated coding signals to generate multiple precoded signals; performs power constraints on the multiple precoded signals to obtain transmission signals, and transmits the transmission signals to the receiver.

[0126] S302 The receiver obtains a received signal transmitted from a target channel, and decodes the received signal based on an automatic decoder to obtain decoded data.

[0127] The transmitter and receiver system of the present application, such as Figure 4 As shown in Figure 2, a dual-user MIMO broadcast channel with confidential messages is taken as an example. Figure 4 The transmitter, receiver 1 and receiver 2 are represented by Tx, Rx1 and Rx2 respectively, and are equipped with n t ,n1 and n2 antennas, where n t ≥n1+n2. The transmitter sends two confidential messages W1 and W2 to two receivers respectively. In order to ensure the confidentiality of the message, the transmitter converts the confidential message vector Encoded into a transmission signal with a block length of n Where R i It is W i The number of bits of the elements in . Receiver i can only decode the corresponding confidential message W i , but cannot decode another message. and is the complex channel gain matrix from the transmitter to receiver 1 and receiver 2, which can be expressed as:

[0128]

[0129] where i∈{1,2}, L i is the number of propagation paths. i,l is the complex channel gain of the lth path, which obeys the complex Gaussian distribution. is the angle at which the lth path reaches the receiver. l is the departure angle of the lth path at the transmitter. Both the arrival angle and the departure angle obey linear distribution. is the receiving array response vector at the i-th receiver, a Tx (·) is the transmit array response vector at the transmitter. When a uniform linear array is used, the response vector of the antenna array is calculated as follows:

[0130]

[0131] Among them, N k, k∈{Tx,Rx1,Rx2} represents the number of antennas of the transmitter, receiver 1 and receiver 2. Where d is the distance between two adjacent antennas in the array, and λ is the carrier wavelength.

[0132] Assume that the confidential messages of two users are transmitted in the same time slot and both the transmitter and the receiver have instantaneous channel state information (CSI). The signals received at receiver 1 and receiver 2 are and They are:

[0133]

[0134] Y2=H2X+z2

[0135] in, is additive white Gaussian noise. The transmitted signal X has an average power constraint:

[0136]

[0137] A strong confidentiality criterion is used to evaluate the security of confidential message transmission. When i≠j, That is, when the block length n→∞, the signal of the unintended receiver should be statistically independent of the corresponding confidential message. To measure the reliability of message transmission, Denotes the decoder for the message W i The estimated value obtained after decoding.

[0138] The information sender of this application converts the information to be sent into a one-hot encoding form. Then, the coded signal is input into the automatic encoder for modulation and coding. Then, the modulated and coded signal is transmitted to the null space of the channel matrix using the zero-forcing precoding technology to generate a precoded signal. After the precoded signal is processed by the power constraint layer, the final transmission signal is formed. At the receiving end, the received signal is decoded by the automatic decoder to restore the original information. This method uses an automatic codec to ensure the reliability of information transmission, and uses the zero-forcing precoding technology to achieve secure communication between multiple users, thereby ensuring the secure transmission of information between nodes in the communication system.

[0139] According to another aspect of an embodiment of the present application, a wireless communication secure transmission device based on deep learning is provided, comprising:

[0140] The transmitter is used to perform one-hot encoding on the message to be sent on each channel to obtain multiple one-hot encoding matrices; input the one-hot encoding matrix and the channel characteristic sequence into the corresponding automatic encoder to obtain multiple modulated coded signals; perform zero-forcing precoding on the multiple modulated coded signals to generate multiple precoded signals; perform power constraints on the multiple precoded signals to obtain transmission signals, and transmit the transmission signals to the receiver;

[0141] The receiver is used to obtain a received signal transmitted from a target channel, and decode the received signal based on an automatic decoder to obtain decoded data.

[0142] It should be noted that the wireless communication security transmission device based on deep learning provided in the above embodiment only uses the division of the above functional modules as an example when executing the wireless communication security transmission method based on deep learning. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the wireless communication security transmission device based on deep learning provided in the above embodiment and the wireless communication security transmission method based on deep learning belong to the same concept. The implementation process is detailed in the method embodiment, which will not be repeated here.

[0143] According to another aspect of an embodiment of the present application, an electronic device corresponding to the deep learning-based wireless communication security transmission method provided in the aforementioned embodiment is also provided to execute the above-mentioned deep learning-based wireless communication security transmission method.

[0144] Please refer to Figure 6 , which shows a schematic diagram of an electronic device provided by some embodiments of the present application. Figure 6 As shown, the electronic device includes: a processor 600, a memory 601, a bus 602 and a communication interface 603, and the processor 600, the communication interface 603 and the memory 601 are connected through the bus 602; the memory 601 stores a computer program that can be run on the processor 600, and when the processor 600 runs the computer program, it executes the deep learning-based wireless communication security transmission method provided in any of the aforementioned embodiments of the present application.

[0145] The memory 601 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 603 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used.

[0146] The bus 602 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. Among them, the memory 601 is used to store programs, and the processor 600 executes the program after receiving the execution instruction. The wireless communication security transmission method based on deep learning disclosed in any implementation of the aforementioned embodiment of the present application can be applied to the processor 600, or implemented by the processor 600.

[0147] The processor 600 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 600. The above processor 600 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiments of the present application can be directly embodied as a hardware decoding processor to be executed, or the hardware and software modules in the decoding processor can be executed. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 601, and the processor 600 reads the information in the memory 601 and completes the steps of the above method in combination with its hardware.

[0148] The electronic device provided in the embodiment of the present application and the wireless communication security transmission method based on deep learning provided in the embodiment of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, operated or implemented therein.

[0149] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0150] The above embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention patent shall be subject to the attached claims.

Claims

1. A wireless communication security transmission method based on deep learning, used in a transmitter, characterized in that: include: Perform one-hot encoding on the messages to be sent on each channel to obtain multiple one-hot encoding matrices; Inputting the one-hot coding matrix and the channel feature sequence into a corresponding automatic encoder to obtain a plurality of modulated coded signals; Performing zero-forcing precoding on the multiple modulated coded signals to generate multiple precoded signals; Power constraints are performed on the multiple precoded signals to obtain a transmit signal, and the transmit signal is transmitted to a receiver.

2. The method according to claim 1, characterized in that Before inputting the one-hot encoding matrix and the channel feature sequence into the corresponding autoencoder, the method further includes: The autoencoder is trained based on a fully connected neural network to obtain the trained autoencoder.

3. The method according to claim 2, characterized in that Training the autoencoder based on a fully connected neural network includes: Randomly generate a preset number of channel matrices: Initialize the autoencoder network, including the transmitter and multiple receiver sub-networks; Randomly generate confidential messages and generate encoded transmission signals based on the transmitter network function; Transmitting the transmission signal to a receiver through a channel layer, and adding Gaussian white noise at the receiver; Using an automatic decoder to decode the estimated confidential message based on the received signal and the channel characteristic sequence; The transmitter and multiple receiver sub-networks are trained simultaneously until the model converges to obtain a trained autoencoder.

4. The method according to claim 2, characterized in that: Also includes: Based on average cross entropy as the loss function for each receiver; Based on the sum of the loss functions of multiple receivers, the total loss function is obtained.

5. The method according to claim 1, characterized in that Power constraint is performed on the multiple precoded signals to obtain a transmission signal, comprising: Introducing power splitting factors to distribute the total power to multiple receivers; Calculating a power limit factor for each precoded signal based on the divided power, and scaling each precoded signal based on the power limit factor; Based on the sum of the scaled precoded signals, a transmit signal is obtained.

6. A wireless communication security transmission method based on deep learning, used in a receiver, characterized in that: include: Obtain the received signal transmitted from the target channel; The received signal is decoded based on an automatic decoder to obtain decoded data.

7. The method according to claim 6, characterized in that After getting the decoded data, it also includes: Input the decoded data into a preset activation function to obtain an output probability matrix; Based on the probability matrix, the codeword with the highest probability is selected as the final output.

8. A wireless communication security transmission method based on deep learning, characterized in that: include: The transmitter performs one-hot encoding on the message to be sent on each channel to obtain multiple one-hot encoding matrices; Inputting the one-hot coding matrix and the channel feature sequence into a corresponding automatic encoder to obtain a plurality of modulated coded signals; Performing zero-forcing precoding on the multiple modulated coded signals to generate multiple precoded signals; Performing power constraints on the multiple precoded signals to obtain a transmit signal, and transmitting the transmit signal to a receiver; The receiver acquires a received signal transmitted from a target channel, and decodes the received signal based on an automatic decoder to obtain decoded data.

9. A wireless communication security transmission device based on deep learning, characterized in that: include: A transmitter, configured to perform one-hot encoding on the messages to be sent on each channel to obtain a plurality of one-hot encoding matrices; Inputting the one-hot coding matrix and the channel feature sequence into a corresponding automatic encoder to obtain a plurality of modulated coded signals; Performing zero-forcing precoding on the multiple modulated coded signals to generate multiple precoded signals; Performing power constraints on the multiple precoded signals to obtain a transmit signal, and transmitting the transmit signal to a receiver; The receiver is used to obtain a received signal transmitted from a target channel, and decode the received signal based on an automatic decoder to obtain decoded data.

10. An electronic device, characterized in that: It includes a processor and a memory storing program instructions, and the processor is configured to execute the wireless communication security transmission method based on deep learning as described in any one of claims 1 to 8 when executing the program instructions.

Images