A semantic camouflage communication method, system, power grid and electronic device

Abstract

The application discloses a semantic camouflage communication method and system, a power grid and electronic equipment, and relates to the technical field of power system safety. The method obtains semantic information extracted from an image to be sent by a semantic encoder of a semantic communication system at a sending end of the semantic communication system, adds artificial noise for ensuring that a legal user and a potential eavesdropper in a deployment environment of the semantic communication system can both decode, and the legal user decodes to obtain the image to be sent, while the potential eavesdropper decodes to obtain a camouflage image, obtains transmission semantic information, pre-encodes the transmission semantic information, and sends the transmission semantic information. Thus, the problem of low data transmission security caused by the fact that an eavesdropper is easily aware of protection measures and adopts more complex attacks when maintaining confidentiality through encryption or a physical layer security system during power data transmission by using semantic communication is solved.

Description

Technical Field

[0001] This invention relates to the field of power system security technology, and more specifically, to a semantic spoofing communication method, system, power grid, and electronic equipment. Background Technology

[0002] During power grid inspections, to ensure the stable operation and efficient management of the power system, it is necessary to collect and transmit large amounts of image data from sensors in real time. This data includes, but is not limited to, equipment operating status monitoring and environmental monitoring information. However, with the development of smart grid technology, power grid inspection systems are becoming increasingly complex, and the amount of image data involved is growing exponentially, placing significant transmission pressure on network infrastructure. To address this challenge, an advanced communication method—semantic communication—is commonly used to alleviate the demand on network bandwidth. Semantic communication focuses not on data transmission but on understanding the meaning of the data and ensuring that the receiving end accurately grasps the sender's intent. Therefore, this approach can significantly reduce the amount of data actually transmitted, thereby lowering the requirements for network bandwidth.

[0003] Understandably, semantic communication requires all participants to possess the same knowledge bases (KBs) and semantic codec network. However, in widely deployed semantic communication systems, the inherent security of semantic communication can be compromised. This occurs because the widespread deployment of semantic encoders and decoders on end devices provides eavesdroppers with potential opportunities to access the legitimate recipient's decoders and KBs. Therefore, eavesdropping attacks in semantic communication become a critical security issue that requires careful consideration before large-scale deployment of semantic communication systems. Many studies have proposed semantic communication approaches that consider security aspects, relying on cryptographic or physical layer security systems to maintain confidentiality; however, these methods lack stealth, allowing eavesdroppers to become aware of information protection measures. Furthermore, eavesdroppers can employ more sophisticated attacks, such as accessing KBs and semantic codec networks, to extract semantic information. Summary of the Invention

[0004] The purpose of this invention is to provide a semantic spoofing communication method, system, power grid, and electronic device to solve the problem that when using semantic communication for power data transmission, maintaining confidentiality through encryption or physical layer security systems makes it easy for eavesdroppers to become aware of the protection measures and launch more sophisticated attacks, resulting in low data transmission security.

[0005] In a first aspect, the present invention provides a semantic spoofing communication method, applied to the sending end of a semantic communication system, the method comprising:

[0006] Obtain semantic information, which is information extracted from the image to be sent by the semantic encoder of the semantic communication system;

[0007] Artificial noise is added to the semantic information to obtain transmission semantic information; the artificial noise is used to ensure that both legitimate users and potential eavesdroppers in the deployment environment of the semantic communication system can decode the transmission semantic information. The legitimate user decodes the transmission semantic information to obtain the image to be sent, and the potential eavesdropper decodes the transmission semantic information to obtain the disguised image.

[0008] The transmitted semantic information is pre-encoded and then sent.

[0009] In a preferred embodiment, obtaining semantic information includes:

[0010] Obtain the image to be sent;

[0011] The image to be sent is input into the semantic encoder of the semantic communication system to extract semantic information and obtain the semantic information.

[0012] The semantic encoder is jointly trained with the semantic decoder of the semantic communication system based on the channel state information of legitimate users and potential eavesdroppers in the deployment environment of the semantic communication system.

[0013] In a preferred embodiment, adding artificial noise to the semantic information includes:

[0014] Obtain the camouflaged image;

[0015] The camouflaged image is input into the artificial noise generation model to obtain the artificial noise output by the artificial noise generation model;

[0016] The artificial noise is applied to the semantic information;

[0017] The artificial noise generation model is a neural network model obtained by co-training with the semantic encoder and the semantic decoder.

[0018] In a preferred embodiment, the artificial noise generation model comprises a two-layer interconnected network;

[0019] The first layer network is an encoder that mimics a semantic communication framework, used to ensure the retrieval of equivalent semantic information of the camouflaged image;

[0020] The second layer network is a U-Net architecture, used to receive the camouflaged image and generate the artificial noise with the same dimensions as the camouflaged image.

[0021] In a preferred embodiment, the artificial noise generation model enables the potential eavesdropper to decode the actual information R. e Tag M near the camouflaged image t The actual information R obtained by the legitimate user through decoding. b The tag M near the image to be sent b ′, and R e Not decoded as M b Construct a loss function L for the objective:

[0022]

[0023] in, This represents the cross-entropy loss.

[0024] In a preferred embodiment, the artificial noise obeys L... ∞ Norm constraints.

[0025] In a preferred embodiment, based on a preset formula, the artificial noise is used to ensure that both the legitimate user and the potential eavesdropper can decode the transmitted semantic information, and the legitimate user decodes to obtain the image to be sent, while the potential eavesdropper decodes to obtain the disguised image;

[0026] The preset formula is expressed as follows:

[0027] arg Distance[f(x p ,G),M t ]

[0028]

[0029] d(x p ,x)≤∈,

[0030] Where f(·) represents the semantic decoder, and d(·,·) represents the expression used to compute L ∞ The norm function, ∈ denotes a pre-specified threshold, d(x) p ,x)≤∈ represents L ∞ Norm constraints.

[0031] In a preferred embodiment, the precoding and transmission of the transmitted semantic information includes:

[0032] Based on the channel state information of the legitimate users, a precoding vector is generated;

[0033] Based on the precoding vector, the transmitted semantic information is precoded;

[0034] The pre-coded semantic information is transmitted via a wireless channel.

[0035] Secondly, the present invention also provides another semantic spoofing communication method, applied to a target receiving end of a semantic communication system, wherein the target receiving end is a receiving end of a legitimate user in the deployment environment of the semantic communication system, and the method includes:

[0036] Acquire transmission semantic information, wherein the transmission semantic information is semantic information with added artificial noise, and the semantic information is information extracted from the image to be sent by the semantic encoder of the semantic communication system;

[0037] The transmitted semantic information is input into the semantic decoder of the semantic communication system for decoding to obtain the image to be sent.

[0038] The semantic encoder and the semantic decoder are jointly trained based on the channel state information of legitimate users and potential eavesdroppers in the deployment environment.

[0039] Thirdly, the present invention also provides another semantic spoofing communication method, applied to a spoofing receiver in a semantic communication system, wherein the spoofing receiver is a receiver of a potential eavesdropper in the deployment environment of the semantic communication system, and the method includes:

[0040] Acquire transmission semantic information, wherein the transmission semantic information is semantic information with added artificial noise, and the semantic information is information extracted from the image to be sent by the semantic encoder of the semantic communication system;

[0041] The transmitted semantic information is input into the semantic decoder of the semantic communication system for decoding, thereby obtaining the camouflaged image output by the semantic decoder;

[0042] The semantic encoder and the semantic decoder are jointly trained based on the channel state information of legitimate users and potential eavesdroppers in the deployment environment of the semantic communication system.

[0043] In a preferred embodiment, the artificial noise is generated by an artificial noise generation model, which is a neural network model obtained by co-training with the semantic encoder and the semantic decoder.

[0044] Fourthly, the present invention provides a semantic spoofing communication system, comprising:

[0045] The first acquisition unit is used to acquire semantic information, which is information extracted from the image to be sent by the semantic encoder of the semantic communication system.

[0046] The first processing unit adds artificial noise to the semantic information to obtain transmission semantic information; the artificial noise is used to ensure that both legitimate users and potential eavesdroppers in the deployment environment of the semantic communication system can decode the transmission semantic information, the legitimate user decodes the transmission semantic information to obtain the image to be sent, and the potential eavesdropper decodes the transmission semantic information to obtain the disguised image.

[0047] The sending unit is used to pre-encode the transmitted semantic information and send it.

[0048] Fifthly, the present invention also provides another semantic spoofing communication system, comprising:

[0049] The second acquisition unit is used to acquire transmission semantic information, which is semantic information with added artificial noise, and the semantic information is information extracted from the image to be sent by the semantic encoder of the semantic communication system.

[0050] The second processing unit is used to input the transmitted semantic information into the semantic decoder of the semantic communication system for decoding to obtain the image to be sent.

[0051] The semantic encoder and the semantic decoder are jointly trained based on the channel state information of legitimate users and potential eavesdroppers in the deployment environment of the semantic communication system.

[0052] Sixthly, the present invention also provides another semantic spoofing communication system, comprising:

[0053] The third acquisition unit is used to acquire transmission semantic information, which is semantic information with added artificial noise, and the semantic information is information extracted from the image to be sent by the semantic encoder of the semantic communication system.

[0054] The third processing unit is used to input the transmitted semantic information into the semantic decoder of the semantic communication system for decoding to obtain the camouflaged image.

[0055] The semantic encoder and the semantic decoder are jointly trained based on the channel state information of legitimate users and potential eavesdroppers in the deployment environment of the semantic communication system.

[0056] In a seventh aspect, the present invention also provides a power grid, including a power system body and a semantic spoofing communication system provided in the fourth, fifth or sixth aspect of the present invention.

[0057] In a seventh aspect, the present invention provides an electronic device, the electronic device comprising:

[0058] processor;

[0059] Memory used to store the processor's executable instructions;

[0060] The processor is configured to execute the semantic spoofing communication method provided in the first, second, or third aspect of the present invention.

[0061] To achieve the above objectives, the semantic spoofing communication method, system, power grid, and electronic equipment provided by this invention, at the transmitting end of the semantic communication system, acquires semantic information extracted from the image to be transmitted by the semantic encoder of the semantic communication system. Then, artificial noise is added to this semantic information to ensure that both legitimate users and potential eavesdroppers in the deployment environment of the semantic communication system can decode it. Legitimate users decode the image to be transmitted, while potential eavesdroppers decode the spoofed image. This yields the transmission semantic information, which is then pre-encoded and transmitted. This achieves the goal of misleading potential eavesdroppers with artificial noise while preserving accurate decoding results for legitimate users. This avoids the drawbacks of eavesdroppers realizing the protection measures and employing more complex attacks, effectively improving the security of power data transmission via semantic communication technology. Attached Figure Description

[0062] Figure 1 This is a flowchart of a semantic spoofing communication method provided in an embodiment of the present invention.

[0063] Figure 2 This is a schematic diagram illustrating how a convolutional layer equivalently represents a precoding and a wireless channel, as provided in an embodiment of the present invention.

[0064] Figure 3 This is a structural diagram of a semantic communication network architecture example provided in an embodiment of the present invention.

[0065] Figure 4 This is a structural diagram of an example of an artificial noise generation model provided in an embodiment of the present invention.

[0066] Figure 5 A comparison diagram of the ordinary semantic communication method and the semantic spoofing communication method provided in the embodiments of the present invention.

[0067] Figure 6 A flowchart of another semantic spoofing communication method provided in an embodiment of the present invention.

[0068] Figure 7 A flowchart of another semantic spoofing communication method provided in an embodiment of the present invention.

[0069] Figure 8 This is a structural diagram of a semantic spoofing communication system provided in an embodiment of the present invention.

[0070] Figure 9This is a structural diagram of another semantic spoofing communication system provided in an embodiment of the present invention.

[0071] Figure 10 This is a structural diagram of another semantic spoofing communication system provided in an embodiment of the present invention.

[0072] Figure 11 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0073] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0074] It should be noted that an end-to-end semantic communication system needs to include a semantic encoder and a semantic decoder, so that the system can semantically encode information at the sending end based on the content of the knowledge base, and then semantically decode it at the receiving end.

[0075] Specifically, the process of training an end-to-end semantic communication system generally includes three steps: communication modeling, precoding and wireless channel transmission, and semantic communication framework training.

[0076] Specifically, regarding the communication modeling process, it is assumed that the semantic communication system includes N t The transmitting end of each antenna, N r Legitimate users of each antenna and those with N e Potential eavesdroppers with multiple antennas. Further assuming the eavesdropping channel follows Rayleigh fading, the channel from the transmitter to the eavesdropper can be defined as... The channel between the sender and the legitimate user can be represented as... It follows Ricean fading. Let x represent the signal of the image message M to be transmitted, encoded by joint semantic coding and channel coding. The precoding matrix is ​​composed of... This indicates that the signals received by the legitimate user and the potential eavesdropper are respectively:

[0077] y b =HFx+n b (1)

[0078] y e =GFx+n e (2)

[0079] Where, n b and n e It is obedience and Additive white Gaussian noise (AWGN). Understandably, in this scenario, due to the widespread deployment of semantic communication systems, both the semantic encoder and decoder are easily accessible to potential eavesdroppers. Therefore, potential eavesdroppers possess the ability to decode the correct information. That is, when both the legitimate user and the potential eavesdropper receive the corresponding signal y... b and y e Then, the same semantic decoder can be used to perform channel decoding and semantic decoding on the signal to obtain their respective decoding results M. b ′ and M e ′, and there exists M b ′ and M e The same possibility.

[0080] Regarding the precoding and wireless channel transmission process, such as Figure 2 As shown, assuming convolutional layers are used to simulate precoding and wireless channel effects, they are defined by the precoding layer and the wireless channel layer, respectively.

[0081] Specifically, This represents the m-th component of the n-th row of F, where m and n are in the range (1, 2, ..., N). t ).element Let H represent the k-th component in the r-th row, where r∈(1,2,...,N). r The signal received by the r-th antenna in the l-th time period is denoted as... Then, through the wireless channel layer, a size of 1×2×N can be obtained. r The feature map of L is then reshaped into 2N. r ×1×L, which represents the signal received by the k-th user at the receiver. The precoding layer contains 2N... t One convolutional kernel, while the wireless channel layer consists of 2N r It consists of several convolutional kernels, thus reflecting the inherent complexity of the physical processes in the semantic-aware communication system.

[0082] This application aims to address the aforementioned problems by enhancing the security of semantic communication during power grid inspections by leveraging the channel differences between legitimate users and potential eavesdroppers from a physical layer security perspective. Specifically, it involves adding semantic-level artificial noise to the semantic information and then exploiting the differences in channels between legitimate users and potential eavesdroppers to mislead the decoding results of potential eavesdroppers, making them unaware of the error, while simultaneously preserving accurate decoding for legitimate users, thereby achieving the goal of "concealing the truth and revealing the false."

[0083] It should be noted that the semantic spoofing communication method provided by the present invention is executed on an electronic device, which may be a smart terminal device such as a laptop, personal computer and tablet computer, or a server on the network side.

[0084] like Figure 1 As shown, the semantic spoofing communication method provided in this embodiment of the invention is applied to the sending end of a semantic communication system and mainly includes the following steps:

[0085] 110. Obtain semantic information.

[0086] The semantic information is the information extracted from the image to be sent by the semantic encoder of the semantic communication system.

[0087] Specifically, before performing semantic spoofing communication, it is necessary to first train the semantic encoder and semantic decoder of the semantic communication system, and this training process needs to be jointly trained with the knowledge base and channel state information.

[0088] Furthermore, after training the semantic encoder and semantic decoder, the sending end of the semantic communication system uses the semantic encoder to extract semantic information from the image to be sent, thereby obtaining semantic information.

[0089] Furthermore, in an optional embodiment, obtaining semantic information includes:

[0090] Get the image to be sent;

[0091] The image to be sent is input into the semantic encoder of the semantic communication system to extract semantic information.

[0092] The semantic encoder is jointly trained with the semantic decoder of the semantic communication system based on the channel state information of legitimate users and potential eavesdroppers in the deployment environment of the semantic communication system.

[0093] Specifically, by accurately estimating the wireless communication channels in the deployment environment, channel state information for legitimate users and potential eavesdroppers can be obtained. By jointly training the semantic encoder and semantic decoder with the channel state information, the semantic encoder and semantic decoder can not only learn key features from the knowledge base but also absorb channel state information, thereby achieving more efficient information transmission.

[0094] In other words, through joint training, the semantic encoder can automatically adjust to adapt to channel characteristics when encoding information. Thus, even if the signal is affected by channel noise during transmission, the semantic decoder can accurately recover the original information, thereby minimizing information loss and errors. This optimizes the performance of the semantic communication system and improves the security and reliability of information transmission.

[0095] 120. Add artificial noise to the semantic information to obtain the transmitted semantic information.

[0096] Artificial noise is used to ensure that both legitimate users and potential eavesdroppers in the deployment environment of the semantic communication system can decode and transmit semantic information. Legitimate users decode and transmit semantic information to obtain the image to be sent, while potential eavesdroppers decode and transmit semantic information to obtain a disguised image.

[0097] In an optional embodiment, artificial noise is added to the semantic information, including:

[0098] Obtain the camouflaged image;

[0099] The camouflaged image is input into the artificial noise generation model to obtain the artificial noise output by the artificial noise generation model;

[0100] Add artificial noise to semantic information;

[0101] The artificial noise generation model is a neural network model obtained by co-training with the semantic encoder and semantic decoder.

[0102] In this embodiment, artificial noise is generated using a disguised image designed to mislead potential eavesdroppers. This artificial noise is then applied to semantic information to obtain transmitted semantic information. This allows legitimate users in the semantic communication system deployment environment to receive the semantic information using a receiver and decode the original information using a semantic decoder. However, when a potential eavesdropper receives the semantic information using a receiver and decodes it using a semantic decoder, the presence of semantic-level artificial noise results in the eavesdropper's decoding result being the disguised information rather than the original information.

[0103] In some possible embodiments, the artificial noise generation model can be a neural network, denoted as Perturbation, which receives an image M intended to mislead a potential eavesdropper. t As input, the output is the artificial noise δ, and the formula can be summarized as follows:

[0104] δ = Perturbation(M) t (3)

[0105] The artificial noise generation network is trained in conjunction with the semantic encoder and semantic decoder.

[0106] Semantic extraction uses the aforementioned semantic encoder to extract semantic information from the image to be sent, i.e., the original information:

[0107] x = Encoder(M) (4)

[0108] Where Encoder is the semantic encoder, M is the original information, and x is the semantic information extracted by the semantic encoder from the original information.

[0109] The generated artificial noise is then applied to the semantic information to obtain the transmitted semantic information:

[0110] x p =Perturbation(M t )+Encoder(M)=δ+x (5)

[0111] 130. Pre-encode the semantic information to be transmitted and then send it.

[0112] Specifically, in an optional embodiment, precoding and transmitting the transmitted semantic information includes:

[0113] Generate a precoding vector based on the channel state information of legitimate users;

[0114] Pre-encode the transmitted semantic information based on the pre-coding vector;

[0115] The pre-coded semantic information is transmitted via a wireless channel.

[0116] Specifically, the transmitting end generates a precoding vector based on the channel state information of the legitimate user, uses the vector to precode the transmission semantic information, and then transmits it through a wireless channel using a transmitter.

[0117] In some possible embodiments, f and h represent the precoding layer and the wireless channel layer, respectively. Here, s represents the data to be transmitted, x represents the precoded data (which has the same structure as s), and z corresponds to the data received by the receiver, i.e., legitimate users and potential eavesdroppers. In semantic communication systems, to avoid complex multiplication caused by complex numbers during the training phase, two convolutional layers are used to equivalently represent the precoding and wireless channel for easier training. The channel s during the transmission phase is divided into L time slots, with each time slot transmitting N... t If there are multiple complex symbols, then the signal transmitted by the user in the Lth time slot of the t-th antenna is:

[0118]

[0119] Where, l∈(1, 2, ..., L), t∈(1, 2, ..., N) t In the superscript, index "1" represents the real part, and "2" represents the imaginary part. In this case, the transmitted signal can be considered to have a shape of 1×2×N. t The feature map of L.

[0120] The original information M is processed by semantic information extraction at the sending end and artificial noise is added to obtain semantic information x. p Signal x pAfter preprocessing and wireless channel transmission, the signal received by the receiver is denoted as z. The receiver, i.e., the legitimate user and the potential eavesdropper, uses their respective decoders for decoding, i.e.:

[0121] M d =Decoder(z) (7)

[0122] As can be understood, as mentioned earlier, the process of training an end-to-end semantic communication system generally includes three steps: communication modeling, precoding and wireless channel transmission, and semantic communication framework training. The artificial noise generation model, on the other hand, is an additional component of the semantic communication framework's transmitting end, used to protect the image to be transmitted without affecting the framework's operation.

[0123] In some alternative embodiments, a semantic segmentation dataset can be used to train the semantic communication framework. For example, this dataset could consist of 1569 urban scene images, segmented into 9 classes for semantic segmentation. For the semantic communication framework, training parameters are set with a batch size of 12, 300 training epochs, and stochastic gradient descent as the deep learning optimizer with an initial learning rate of 0.02. The momentum coefficient is set to 0.9 to enhance optimization stability, while the L2 regularization coefficient (weight decay) is kept at 0.0005 to mitigate overfitting. Furthermore, a learning rate decay of 0.01 is applied, where the learning rate decreases to 0.99 of its original value in each training iteration. The loss function used is cross-entropy loss.

[0124] The model structure is as follows Figure 3 As shown, it should be noted that in the semantic encoder, unless otherwise specified, all convolutional layers are identical, i.e., a kernel size of 3×3, a stride of 1, padding of 1, and some max-pooling is added between convolutional layers for subsampling. Each Mini-inception module contains two parallel convolutional layers: one with a kernel size of 3×3, a stride of 1, padding of 1, and dilation of 1, and the other with a dilation of 2. The output of the Mini-inception is the concatenation of the outputs of these two convolutional layers. The number of output channels on each convolutional layer in the Mini-inception module is equal to half the number of input channels of the mini-start block. Furthermore, each convolutional layer in the semantic encoder and semantic decoder is followed by a batch normalization layer, then a non-linear activation function Leaky-ReLU. In the last three convolutional layers, the input image is compressed, shaped, and normalized to meet signal power constraints. Finally, the output of the semantic encoder is the complex-valued wireless channel input symbol.

[0125] In the semantic decoder, since the number of antennas at the receiver may differ from the number of antennas at each transmitter, a post-processing layer can be used to ensure that the dimension of the received complex-valued signal is the same as that of the transmitted symbols. The input size of the semantic decoder is N. r ×2×1200 / N t Assume N r It can be N t Divisibility, the post-processing layer is a convolutional layer with a kernel size of 2×N. r The step size is 2×N r Filled with 0, it has 2N t The semantic decoder uses a convolutional kernel followed by batch normalization and a Leaky-ReLU activation function. Next, convolutional layers are used to recover the channels, and unpooling layers are used to recover the size. The output of the semantic decoder then contains nine categories.

[0126] Furthermore, with the help of the trained semantic communication framework, a deep neural network for generating artificial noise can be trained.

[0127] Understandably, the network needs to be jointly trained with the knowledge base and the eavesdropping channel based on feedback from the semantic communication framework.

[0128] Specifically, the process of training a deep neural network for generating artificial noise includes two steps: problem definition and model building and training.

[0129] Regarding the definition of the problem, in order to enhance protection against eavesdropping threats without compromising communication efficiency, the sending end can use, for example... Figure 4 The illustrated artificial noise generation model, Perturbation, generates semantic-level artificial noise, which is then added to the semantic information to be transmitted. This method can blur any original image and provides defense against eavesdropping attacks without requiring a retraining process.

[0130] That is: given x p =δ+x, where x p Substituting into the communication model, the transmission results for legitimate users are as follows:

[0131]

[0132] Where (HFx+n b HFδ is the result of the semantic information to be transmitted after continuous encoding and channel transmission. HFδ is considered as artificial noise controlled by the variable δ, where the channel matrix H and the precoding matrix F are known. Similarly, the signal at the potential eavesdropper's location is as follows:

[0133]

[0134] By observing the formulas for legitimate users and potential eavesdroppers, it can be found that controlling the values ​​of artificial noise HFδ and GFδ by controlling the value of δ can confuse potential eavesdroppers, and this process can be achieved through deep neural networks.

[0135] Specifically, by specifying the optimization direction of the model using a loss function, the artificial noise GFδ can be transformed into a perturbation targeting the semantic information (GFx+n) using generative adversarial perturbation techniques. e The resistive perturbation of the semantic information guides the decoding result onto the disguised image. Therefore, the problem of defensive adversarial attacks on semantic communication systems in the presence of potential eavesdroppers can be expressed as follows:

[0136]

[0137] Where f(·) is the semantic decoder in the semantic communication system, and d(·,·) represents the semantic decoder used to compute L ∞ The norm is a function, where ∈ represents a pre-specified threshold, d(x) p ,x)≤∈ represents L ∞ Norm constraints are used to ensure that δ does not affect the correct decoding of x. Because x p =δ+x, therefore the goal is to find a perturbation δ such that when it is added from the target input domain... When given any input image, it will cause the decoder f(·) to misclassify a potential eavesdropper, while a legitimate user can use it for decoding.

[0138] The main difference between legitimate users and eavesdroppers is the difference between H and G. In order to achieve the goal of distorting the decoding results of potential eavesdroppers, the generated artificial noise needs to be able to exploit the difference between eavesdropping and legitimate channel state information (CSI).

[0139] Furthermore, regarding model building and training, the artificial noise generation model Perturbation includes two interconnected networks: an encoder that mimics a semantic communication framework to ensure the retrieval of equivalent semantic information from the masquerading image, and a U-Net architecture that takes input from the encoder and generates artificial noise of the same dimension.

[0140] Furthermore, in practical implementation, given the relatively small amount of semantic information data, the model can maintain a four-layer structure compared to the original U-Net model. In some possible embodiments, the convolutional kernel dimension is adjusted to 3×3, the stride is 1, and the padding is (1,1) to ensure that the output of each layer matches the size of the input.

[0141] Except for the top layers in the semantic encoder and semantic decoder, all layers undergo normalization to reduce overfitting. The Leaky ReLU activation function is used in all layers except the highest layer of the semantic encoder to enhance nonlinear representation. Artificial noise is constrained to obey L... ∞ Norm constraint, where L ∞ The norm refers to the norm of a vector being the absolute value of the element with the largest absolute value among all elements of the vector, thereby minimizing its impact on semantic information and ensuring that it reduces the decoding effect at the expected receiver.

[0142] Meanwhile, the loss function includes the following three terms. The first term ensures that legitimate users decode correctly, the second term ensures that potential eavesdroppers are successfully misled, and the third term ensures that eavesdroppers can decode correctly, even if the result is false. That is, the loss function L is defined as:

[0143]

[0144] Among them, R b and R e These represent the actual information decoded by legitimate users and potential eavesdroppers, respectively. M represents the cross-entropy loss. b ′ represents the label of the image to be sent, while M t ' indicates a label for a disguised image intended to mislead potential eavesdroppers, and the superscript "'" indicates an input field. The image labels are used. The goal of the loss function is to minimize the eavesdropper R. e The actual information decoded is close to M. t ′, while ensuring R b Approaching M b And R e Cannot be decoded as M b ′ b .

[0145] It's important to note that the training settings for the artificial noise generation model are very similar to those for the semantic communication framework. In practice, the training batch size for the artificial noise generation model can be set to 12, with no predefined maximum epoch limit. Training dynamically terminates based on the behavior of the loss function. For example, training stops when the loss function value hasn't decreased after 20 epochs. The optimization algorithm can be stochastic gradient descent (SGD), with an initial learning rate of 0.02, a momentum coefficient of 0.9, and an L2 regularization coefficient (weight decay) of 0.0005, used for stabilizing training and mitigating overfitting, respectively. Furthermore, a learning rate decay of 0.01 can also be applied.

[0146] In summary, the overall architecture of the semantic spoofing communication method provided in the above embodiments of the present invention is as follows: Figure 5 As shown, by Figure 5 As can be seen, this method utilizes the semantic information of camouflaged images to generate artificial noise, achieving the ability to blur any original image using a low-complexity architecture, thus providing defense against semantic communication eavesdropping attacks without requiring a retraining process. On the other hand, it trains a neural network to capture channel characteristics, generating artificial noise with different effects as it passes through different channels. By adding this artificial noise to the semantic information, it employs defensive adversarial attack techniques to resist eavesdropping attacks in semantic communication and mislead the decoding results of potential eavesdroppers, thereby achieving secure communication in the end-to-end semantic communication system during power line inspection.

[0147] In summary, by applying the semantic spoofing communication method provided in this embodiment of the invention, devices can transmit information securely based on defensive anti-attack security technology. This ensures low-bandwidth, high-accuracy communication between the communicating parties while resisting eavesdropping attacks from potential eavesdroppers and misleading their decoding results so that they cannot detect it. In this way, the semantic communication system in the power inspection system can resist eavesdropping attacks.

[0148] like Figure 6 As shown in the figure, this embodiment of the invention also provides another semantic spoofing communication method, applied to the target receiving end of a semantic communication system. The target receiving end is the receiving end of a legitimate user in the deployment environment of the semantic communication system, and mainly includes the following steps:

[0149] 210. Obtain the semantic information of the transmission.

[0150] The transmitted semantic information is semantic information with added artificial noise, and the semantic information is information extracted from the image to be sent by the semantic encoder of the semantic communication system.

[0151] 220. Input the semantic information to be transmitted into the semantic decoder of the semantic communication system for decoding to obtain the image to be sent.

[0152] The semantic encoder and semantic decoder are jointly trained based on channel state information of legitimate users and potential eavesdroppers in the deployment environment.

[0153] like Figure 7 As shown in the figure, this embodiment of the invention also provides another semantic spoofing communication method, applied to a spoofing receiver in a semantic communication system. The spoofing receiver is the receiver of a potential eavesdropper in the deployment environment of the semantic communication system, and mainly includes the following steps:

[0154] 310. Obtain the semantic information of the transmission.

[0155] The transmitted semantic information is semantic information with added artificial noise, and the semantic information is information extracted from the image to be sent by the semantic encoder of the semantic communication system.

[0156] 320. Input the transmitted semantic information into the semantic decoder of the semantic communication system for decoding to obtain the camouflaged image.

[0157] The semantic encoder and semantic decoder are jointly trained based on channel state information of legitimate users and potential eavesdroppers in the deployment environment.

[0158] In an optional embodiment, the artificial noise is generated using an artificial noise generation model.

[0159] The artificial noise generation model is a neural network model obtained by co-training with the semantic encoder and semantic decoder.

[0160] The following describes a semantic spoofing communication system provided by an embodiment of the present invention. The semantic spoofing communication system described below can be considered as a modular architecture for implementing the semantic spoofing communication method provided by the embodiment of the present invention; the following description can be referred to in conjunction with the above.

[0161] Optional, see Figure 8 , Figure 8 This is a structural block diagram of a semantic spoofing communication system provided in an embodiment of the present invention. The system may include:

[0162] The first acquisition unit 10 is used to acquire semantic information, which is information extracted from the image to be sent by the semantic encoder of the semantic communication system.

[0163] The first processing unit 20 adds artificial noise to the semantic information to obtain the transmission semantic information. The artificial noise is used to ensure that both legitimate users and potential eavesdroppers in the deployment environment of the semantic communication system can decode the transmission semantic information. Legitimate users decode the transmission semantic information to obtain the image to be sent, and potential eavesdroppers decode the transmission semantic information to obtain the disguised image.

[0164] The transmitting unit 30 is used to pre-encode the transmitted semantic information and transmit it.

[0165] Optional, see Figure 9 , Figure 9 This is a structural block diagram of another semantic spoofing communication system provided in an embodiment of the present invention. The system may include:

[0166] The second acquisition unit 40 is used to acquire transmission semantic information, which is semantic information with added artificial noise, and the semantic information is information extracted from the image to be sent by the semantic encoder of the semantic communication system.

[0167] The second processing unit 50 is used to input the transmitted semantic information into the semantic decoder of the semantic communication system, decode it, and obtain the image to be sent.

[0168] The semantic encoder and semantic decoder are jointly trained based on the channel state information of legitimate users and potential eavesdroppers in the deployment environment of the semantic communication system.

[0169] Optional, see Figure 10 , Figure 10 This is a structural block diagram of another semantic spoofing communication system provided in an embodiment of the present invention. The system may include:

[0170] The third acquisition unit 60 is used to acquire transmission semantic information, which is semantic information with added artificial noise. The semantic information is information extracted from the image to be sent by the semantic encoder of the semantic communication system.

[0171] The third processing unit 70 is used to input the transmitted semantic information into the semantic decoder of the semantic communication system for decoding to obtain the camouflaged image.

[0172] The semantic encoder and semantic decoder are jointly trained based on the channel state information of legitimate users and potential eavesdroppers in the deployment environment of the semantic communication system.

[0173] Optionally, embodiments of the present invention also provide a power grid, which includes a power system body and a semantic spoofing communication system as provided in any of the above embodiments.

[0174] Below, for reference Figure 11 The electronic device provided in the embodiments of this application can be described as follows: at least one processor 100, at least one communication interface 200, at least one memory 300 and at least one communication bus 400;

[0175] In this embodiment of the invention, the number of processor 100, communication interface 200, memory 300, and communication bus 400 is at least one, and the processor 100, communication interface 200, and memory 300 communicate with each other through communication bus 400; obviously, Figure 11 The communication connections shown for the processor 100, communication interface 200, memory 300, and communication bus 400 are optional.

[0176] Optionally, the communication interface 200 can be an interface of a communication module, such as the interface of a GSM module; the processor 100 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.

[0177] The memory 300 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0178] Specifically, the processor 100 is used to execute the application program in the memory to implement the steps of the crane hoist wire rope installation control method described above.

[0179] The above embodiments are merely illustrative examples and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A semantic spoofing communication method, applied to the sending end of a semantic communication system, characterized in that, The method includes: Obtain semantic information, which is information extracted from the image to be sent by the semantic encoder of the semantic communication system; Artificial noise is added to the semantic information to obtain transmission semantic information; based on a preset formula, the artificial noise is used to ensure that both legitimate users and potential eavesdroppers in the deployment environment of the semantic communication system can decode the transmission semantic information. The legitimate user decodes the transmission semantic information to obtain the image to be sent, and the potential eavesdropper decodes the transmission semantic information to obtain a disguised image. The preset formula is expressed as follows: arg Distance[f(x p ,G),M t ] d(x p ,x)≤∈, Where f(·) represents the semantic decoder, and d(·,·) represents the expression used to compute L ∞ The norm function, ∈ denotes a pre-specified threshold, d(x) p ,x)≤∈ represents L ∞ Norm constraint, x p Let G represent the semantic information after adding artificial noise, and M represent the channel matrix of the potential eavesdropper. t Let x represent the spoofed image, x represent the transmitted semantic information, and H represent the channel matrix of the legitimate user. The transmitted semantic information is pre-encoded and then sent.

2. The semantic spoofing communication method according to claim 1, characterized in that, The acquisition of semantic information includes: Obtain the image to be sent; The image to be sent is input into the semantic encoder of the semantic communication system to extract semantic information and obtain the semantic information. The semantic encoder is jointly trained with the semantic decoder of the semantic communication system based on the channel state information of legitimate users and potential eavesdroppers in the deployment environment of the semantic communication system.

3. The semantic spoofing communication method according to claim 2, characterized in that, Adding artificial noise to the semantic information includes: Obtain the camouflaged image; The camouflaged image is input into the artificial noise generation model to obtain the artificial noise output by the artificial noise generation model; The artificial noise is applied to the semantic information; The artificial noise generation model is a neural network model obtained by co-training with the semantic encoder and the semantic decoder.

4. The semantic spoofing communication method according to claim 3, characterized in that, The artificial noise generation model comprises two interconnected networks. The first layer network is an encoder that mimics a semantic communication framework, used to ensure the retrieval of equivalent semantic information of the camouflaged image; The second layer network is a U-Net architecture, used to receive the camouflaged image and generate the artificial noise with the same dimensions as the camouflaged image.

5. The semantic spoofing communication method according to claim 4, characterized in that, The artificial noise generation model enables the potential eavesdropper to decode the actual information R. e The tag M' near the camouflaged image t The actual information R obtained by the legitimate user through decoding b The tag M' near the image to be sent b And R e Not decoded as M' b Construct a loss function L for the target: in, This represents the cross-entropy loss.

6. The semantic spoofing communication method according to claim 5, characterized in that, The artificial noise obeys L ∞ Norm constraints.

7. The semantic spoofing communication method according to claim 1, characterized in that, The step of pre-encoding and sending the transmitted semantic information includes: Based on the channel state information of the legitimate users, a precoding vector is generated; Based on the precoding vector, the transmitted semantic information is precoded; The pre-coded semantic information is transmitted via a wireless channel.

8. A semantic spoofing communication method, applied to a target receiver of a semantic communication system, wherein the target receiver is a receiver of a legitimate user in the deployment environment of the semantic communication system, characterized in that, The method includes: Acquire transmission semantic information, wherein the transmission semantic information is semantic information with added artificial noise, and the semantic information is information extracted from the image to be sent by the semantic encoder of the semantic communication system; The transmitted semantic information is input into the semantic decoder of the semantic communication system for decoding to obtain the image to be sent. The semantic encoder and the semantic decoder are jointly trained based on channel state information of legitimate users and potential eavesdroppers in the deployment environment. Based on a preset formula, the artificial noise is used to ensure that both the legitimate user and the potential eavesdropper can decode the transmitted semantic information. The legitimate user decodes the transmitted semantic information to obtain the image to be sent, and the potential eavesdropper decodes the transmitted semantic information to obtain the disguised image. The preset formula is expressed as follows: arg Distance[f(x p ,G),M t ] d(x p ,x)≤∈, Where f(·) represents the semantic decoder, and d(·,·) represents the expression used to compute L ∞ The norm function, ∈ denotes a pre-specified threshold, d(x) p ,x)≤∈ represents L ∞ Norm constraint, x p Let G represent the semantic information after adding artificial noise, and M represent the channel matrix of the potential eavesdropper. t Let represent the spoofed image, x represent the transmitted semantic information, and H represent the channel matrix of the legitimate user.

9. A semantic spoofing communication method, applied to a spoofed receiver in a semantic communication system, wherein the spoofed receiver is a receiver of a potential eavesdropper in the deployment environment of the semantic communication system, characterized in that... The method includes: Acquire transmission semantic information, wherein the transmission semantic information is semantic information with added artificial noise, and the semantic information is information extracted from the image to be sent by the semantic encoder of the semantic communication system; The transmitted semantic information is input into the semantic decoder of the semantic communication system for decoding, thereby obtaining the camouflaged image output by the semantic decoder; The semantic encoder and the semantic decoder are jointly trained based on channel state information of legitimate users and potential eavesdroppers in the deployment environment. Based on a preset formula, the artificial noise is used to ensure that both the legitimate user and the potential eavesdropper can decode the transmitted semantic information. The legitimate user decodes the transmitted semantic information to obtain the image to be sent, and the potential eavesdropper decodes the transmitted semantic information to obtain the disguised image. The preset formula is expressed as follows: arg Distance[f(x p ,G),M t ] d(x p ,x)≤∈, Where f(·) represents the semantic decoder, and d(·,·) represents the expression used to compute L ∞ The norm function, ∈ denotes a pre-specified threshold, d(x) p ,x)≤∈ represents L ∞ Norm constraint, x p Let G represent the semantic information after adding artificial noise, and M represent the channel matrix of the potential eavesdropper. t Let represent the spoofed image, x represent the transmitted semantic information, and H represent the channel matrix of the legitimate user.

10. The semantic spoofing communication method according to claim 9, characterized in that, The artificial noise is generated by an artificial noise generation model, which is a neural network model obtained by co-training with the semantic encoder and the semantic decoder.

11. A semantic spoofing communication system, applied to the sending end of a semantic communication system, characterized in that, include: The first acquisition unit is used to acquire semantic information, which is information extracted from the image to be sent by the semantic encoder of the semantic communication system. The first processing unit adds artificial noise to the semantic information to obtain the transmitted semantic information; Based on a preset formula, the artificial noise is used to ensure that both legitimate users and potential eavesdroppers in the deployment environment of the semantic communication system can decode the transmitted semantic information. The legitimate user decodes the transmitted semantic information to obtain the image to be sent, and the potential eavesdropper decodes the transmitted semantic information to obtain the disguised image. The preset formula is expressed as follows: arg Distance[f(x p ,G),M t ] d(x p ,x)≤∈, Where f(·) represents the semantic decoder, and d(·,·) represents the expression used to compute L ∞ The norm function, ∈ denotes a pre-specified threshold, d(x) p ,x)≤∈ represents L ∞ Norm constraint, x p Let G represent the semantic information after adding artificial noise, and M represent the channel matrix of the potential eavesdropper. t Let x represent the spoofed image, x represent the transmitted semantic information, and H represent the channel matrix of the legitimate user. The sending unit is used to pre-encode the transmitted semantic information and send it.

12. A semantic spoofing communication system, applied to a target receiving end of a semantic communication system, wherein the target receiving end is a receiving end of a legitimate user in the deployment environment of the semantic communication system, characterized in that, include: The second acquisition unit is used to acquire transmission semantic information, which is semantic information with added artificial noise, and the semantic information is information extracted from the image to be sent by the semantic encoder of the semantic communication system. The second processing unit is used to input the transmitted semantic information into the semantic decoder of the semantic communication system for decoding to obtain the image to be sent. The semantic encoder and the semantic decoder are jointly trained based on channel state information of legitimate users and potential eavesdroppers in the deployment environment of the semantic communication system. Based on a preset formula, the artificial noise is used to ensure that both the legitimate user and the potential eavesdropper can decode the transmitted semantic information. The legitimate user decodes the transmitted semantic information to obtain the image to be sent, and the potential eavesdropper decodes the transmitted semantic information to obtain the disguised image. The preset formula is expressed as follows: arg Distance[f(x p ,G),M t ] d(x p ,x)≤∈, Where f(·) represents the semantic decoder, and d(·,·) represents the expression used to compute L ∞ The norm function, ∈ denotes a pre-specified threshold, d(x) p ,x)≤∈ represents L ∞ Norm constraint, x p Let G represent the semantic information after adding artificial noise, and M represent the channel matrix of the potential eavesdropper. t Let represent the spoofed image, x represent the transmitted semantic information, and H represent the channel matrix of the legitimate user.

13. A semantic spoofing communication system, applied to a spoofing receiver of a semantic communication system, wherein the spoofing receiver is a receiver of a potential eavesdropper in the deployment environment of the semantic communication system, characterized in that, include: The third acquisition unit is used to acquire transmission semantic information, which is semantic information with added artificial noise, and the semantic information is information extracted from the image to be sent by the semantic encoder of the semantic communication system. The third processing unit is used to input the transmitted semantic information into the semantic decoder of the semantic communication system for decoding to obtain the camouflaged image. The semantic encoder and the semantic decoder are jointly trained based on channel state information of legitimate users and potential eavesdroppers in the deployment environment of the semantic communication system. Based on a preset formula, the artificial noise is used to ensure that both the legitimate user and the potential eavesdropper can decode the transmitted semantic information. The legitimate user decodes the transmitted semantic information to obtain the image to be sent, and the potential eavesdropper decodes the transmitted semantic information to obtain the disguised image. The preset formula is expressed as follows: arg Distance[f(x p ,G),M t ] d(x p ,x)≤∈, Where f(·) represents the semantic decoder, and d(·,·) represents the expression used to compute L ∞ The norm function, ∈ denotes a pre-specified threshold, d(x) p ,x)≤∈ represents L ∞ Norm constraint, x p Let G represent the semantic information after adding artificial noise, and M represent the channel matrix of the potential eavesdropper. t Let represent the spoofed image, x represent the transmitted semantic information, and H represent the channel matrix of the legitimate user.

14. A power grid, characterized in that, This includes the power system itself and the semantic spoofing communication system as described in claim 12 or 13.

15. An electronic device, characterized in that, The electronic device includes: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the semantic spoofing communication method as described in any one of claims 1 to 8 or claim 9 or 10.

Images