Semantic communication system based on cooperative interference

Abstract

The invention discloses a semantic communication system based on cooperative interference. The system comprises a signal transmitting end, a signal receiving end and a cooperative jammer, the signal sending end is used for acquiring semantic information to be sent and sending the semantic information to be sent to the signal receiving end through the main channel; wherein the main channel is a wireless channel between a signal transmitting end and a signal receiving end; the cooperative jammer is used for generating anti-interference noise when obtaining semantic information to be sent, and sending the anti-interference noise to the signal receiving end through the main channel when the signal sending end sends the semantic information to be sent to the signal receiving end through the main channel; and the signal receiving end is used for receiving a mixed signal containing the to-be-sent semantic information and the anti-interference noise, and extracting to-be-received semantic information corresponding to the to-be-sent semantic information according to the mixed signal. According to the invention, the security and reliability of semantic information transmission through a wireless channel can be improved.

Description

Technical Field

[0001] The present invention relates to the field of semantic communication security technology, and in particular to a semantic communication system based on collaborative interference. Background Art

[0002] With the continuous development of wireless communication technology, semantic communication has become a research hotspot for the next generation of communication systems due to its efficient bandwidth utilization and robustness to channel imperfections. However, the transmission mode of semantic communication wireless channels also brings security issues in the process of semantic information transmission. The protection of semantic information transmission by existing semantic communication technologies mainly relies on encrypting the required transmitted semantic information or processing it with channel coding technology to protect the required transmitted semantic information. However, this method is difficult to effectively resist the attack of eavesdroppers when facing eavesdropping. Summary of the invention

[0003] An embodiment of the present invention provides a semantic communication system based on cooperative interference. Through cooperative interference, anti-interference noise is added when semantic information to be sent is sent, so that the semantic information to be sent and the anti-interference noise exist in the main channel at the same time. The anti-interference noise confuses the eavesdropper, making it difficult for the eavesdropper to directly obtain the semantic information to be sent, thereby improving the security and transmission reliability of the semantic information transmitted through the wireless channel.

[0004] An embodiment of the present invention provides a semantic communication system based on cooperative interference, comprising: a signal transmitting end, a signal receiving end and a cooperative interference machine;

[0005] The signal transmitting end is used to obtain the semantic information to be sent, and send the semantic information to be sent to the signal receiving end through the main channel; wherein the main channel is a wireless channel between the signal transmitting end and the signal receiving end;

[0006] The cooperative jammer is used to generate anti-interference noise when acquiring semantic information to be sent, and send the anti-interference noise to the signal receiving end through the main channel when the signal sending end sends the semantic information to be sent to the signal receiving end through the main channel;

[0007] The signal receiving end is used to receive a mixed signal including the semantic information to be sent and the anti-interference noise, and extract the semantic information to be received corresponding to the semantic information to be sent according to the mixed signal.

[0008] Further, the signal sending end includes a first encoder;

[0009] The acquiring the semantic information to be sent and sending the semantic information to be sent to the signal receiving end through the main channel includes:

[0010] Obtaining semantic information to be sent;

[0011] Inputting the semantic information to be sent into a first encoder so that the first encoder outputs a semantic coding signal to be sent;

[0012] The semantically coded signal to be sent is sent to a signal receiving end through a main channel.

[0013] Further, the cooperative jammer includes a second encoder;

[0014] The generating of anti-interference noise when acquiring the semantic information to be sent includes:

[0015] Acquire random noise when acquiring semantic information to be sent;

[0016] The random noise is input into a second encoder so that the output of the second encoder is resistant to interference noise.

[0017] Further, the signal receiving end includes a first decoder and a denoiser;

[0018] The extracting the semantic information to be received corresponding to the semantic information to be sent according to the mixed signal includes:

[0019] Inputting the mixed signal into a denoiser so that the denoiser outputs a semantically coded signal to be received corresponding to the semantically coded signal to be sent;

[0020] The semantically coded signal to be received is input into a first decoder, so that the first decoder outputs semantic information to be received.

[0021] Furthermore, the construction of the first encoder and the first decoder includes:

[0022] Acquire a semantic information training sample set; wherein the semantic information training sample set includes a plurality of semantic information samples;

[0023] constructing an initial first encoder and an initial first decoder;

[0024] The semantic information sample is used as the input of the initial first encoder so that the initial first encoder outputs a semantic coding signal sample, and the semantic coding signal sample is used as the input of the initial first decoder so that the initial first decoder outputs a semantic information prediction sample corresponding to the semantic information sample. The initial first encoder and the initial first decoder are jointly trained until the first joint loss function of the initial first encoder and the initial first decoder is minimized, thereby generating a first encoder and a first decoder; wherein, in each joint training process, the loss of the current joint training is determined according to the semantic information sample and the semantic information prediction sample corresponding to the semantic information sample.

[0025] Furthermore, the semantic communication system further comprises: a wiretapping simulator;

[0026] The wiretapping simulator is used to obtain the mixed signal and output a wiretapping simulation signal according to the mixed signal.

[0027] Further, the wiretap simulator includes a second decoder;

[0028] The method outputs a wiretap simulation signal according to the mixed signal, comprising:

[0029] The mixed signal is input into a second decoder so that the second decoder outputs a wiretap simulation signal.

[0030] Furthermore, the construction of the second encoder and the denoiser includes:

[0031] freezing encoder parameters of the first encoder and decoder parameters of the second decoder;

[0032] Obtain semantic information training sample sets and random noise samples;

[0033] Constructing an initial second encoder and an initial denoiser;

[0034] The random noise samples are used as the input of the initial second encoder, the adversarial interference noise prediction samples corresponding to the random noise samples are used as the output of the initial second encoder, the mixed signal samples constructed by the semantic information samples and the adversarial interference noise prediction samples are used as the input of the initial denoiser, the semantic coding prediction signal corresponding to the semantic information samples is used as the output, the mixed signal samples are used as the input of the second decoder, and the eavesdropping simulation prediction signal of the second decoder is used as the output. The initial second encoder and the initial denoiser are jointly adversarially trained until the second joint loss function of the initial second encoder and the initial denoiser is minimized, thereby generating the second encoder and denoiser.

[0035] Furthermore, when the initial second encoder and the initial denoiser are jointly trained in adversarial fashion, encoder parameters of the initial second encoder and denoising parameters of the initial denoiser are adjusted according to the eavesdropping simulation prediction signal, the semantic coding prediction signal and the semantic information sample.

[0036] Furthermore, the semantic information to be sent and the anti-interference noise are sent to the signal receiving end at the same frequency.

[0037] The following beneficial effects are achieved by implementing the present invention:

[0038] The present invention provides a semantic communication system based on cooperative interference, which includes a signal transmitting end, a signal receiving end and a cooperative jammer; after obtaining the semantic information to be sent through the signal transmitting end, the semantic information to be sent is sent to the signal receiving end through the main channel; while the signal transmitting end obtains the semantic information to be sent, the cooperative jammer generates anti-interference noise, and when the signal transmitting end sends the semantic information to be sent to the signal receiving end through the main channel, the anti-interference noise is sent to the signal receiving end through the main channel; when the signal receiving end receives a mixed signal containing the semantic information to be sent and the anti-interference noise, the semantic information to be received corresponding to the semantic information to be sent in the mixed signal. By means of cooperative interference, the anti-interference noise is added when the semantic information to be sent is sent, so that the semantic information to be sent and the anti-interference noise exist in the main channel at the same time, and the anti-interference noise confuses the eavesdropper so that it is difficult for the eavesdropper to directly obtain the semantic information to be sent, thereby improving the security and transmission reliability of the semantic information transmitted through the wireless channel. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a structural diagram of a semantic communication system based on collaborative interference provided by one embodiment of the present invention.

[0040] Figure 2 It is a schematic diagram of a joint training process of a first encoder and a first decoder provided in one embodiment of the present invention.

[0041] Figure 3 It is a schematic diagram of the joint training process of the second encoder and the denoiser provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0044] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" in the specification and claims of this application and the above-mentioned figure descriptions and any variations thereof are intended to cover non-exclusive inclusions.

[0045] In the description of the embodiments of the present application, the technical terms "first", "second", etc. are only used to distinguish different objects, and cannot be understood as indicating or implying relative importance or implicitly indicating the number, specific order or primary and secondary relationship of the indicated technical features. In the description of the embodiments of the present application, the meaning of "multiple" is more than two, unless otherwise clearly and specifically defined.

[0046] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0047] In the description of the embodiments of the present application, the term "and / or" is only a description of the association relationship of the associated objects, indicating that there may be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0048] In the description of the embodiments of the present application, the term "multiple" refers to more than two (including two). Similarly, "multiple groups" refers to more than two groups (including two groups), and "multiple pieces" refers to more than two pieces (including two pieces).

[0049] In the description of the embodiments of the present application, unless otherwise clearly specified and limited, technical terms such as "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, and it can be the internal connection of two elements or the interaction relationship between two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the embodiments of the present application can be understood according to the specific circumstances.

[0050] In order to better understand the technical solution of the present invention, the professional terms involved in the present invention are explained as follows:

[0051] Physical layer security: Physical layer security is a technology that uses the physical characteristics of the communication system to achieve secure communication. Unlike traditional cryptographic methods, physical layer security starts to protect information at the physical layer, using the uncertainty and randomness of the channel to prevent eavesdroppers from obtaining useful information. For example: in image transmission, physical layer security can ensure that even if an eavesdropper intercepts the transmission signal, the original image cannot be restored through methods such as channel coding and signal processing. This method provides additional protection for the confidentiality of the image, especially in wireless communication environments, where the randomness and multipath effects of the channel can be used to enhance the security of information transmission.

[0052] Collaborative interference: Collaborative interference is a technology that improves communication security by introducing controlled interference signals into the network. Interference signals are sent through cooperative nodes to reduce the quality of the received signal for eavesdroppers while minimizing the impact on signal reception and extraction by legitimate receivers. For example, in image transmission, collaborative interference can be performed by cooperative nodes sending specific interference signals to interfere with eavesdroppers through interference signals, making it impossible for eavesdroppers to effectively steal transmitted image data.

[0053] The present invention is described in detail below:

[0054] like Figure 1 As shown, a semantic communication system based on cooperative interference is provided in one embodiment of the present invention, comprising: a signal transmitting end, a signal receiving end and a cooperative jammer;

[0055] The signal transmitting end is used to obtain the semantic information to be sent, and send the semantic information to be sent to the signal receiving end through the main channel; wherein the main channel is a wireless channel between the signal transmitting end and the signal receiving end;

[0056] A cooperative jammer, configured to generate counter-interference noise when acquiring semantic information to be sent, and send the counter-interference noise to the signal receiving end through the main channel when the signal sending end sends the semantic information to be sent to the signal receiving end through the main channel;

[0057] The signal receiving end is used to receive a mixed signal containing the semantic information to be sent and anti-interference noise, and extract the semantic information to be received corresponding to the semantic information to be sent according to the mixed signal.

[0058] Specifically, the semantic information to be sent is obtained at the signal sending end, and the semantic information to be sent is sent to the signal receiving end through the main channel; wherein the semantic information to be sent can be text type semantic information, image type semantic information or sound type semantic information.

[0059] In a preferred embodiment, the signal transmitting end includes a first encoder;

[0060] Acquire semantic information to be sent, and send the semantic information to be sent to a signal receiving end through a main channel, including: acquire semantic information to be sent; input the semantic information to be sent into a first encoder so that the first encoder outputs a semantic coding signal to be sent; send the semantic coding signal to be sent to the signal receiving end through the main channel.

[0061] Specifically, a first encoder is deployed in the signal transmitting end, and the first encoder is used to compress the high-dimensional semantic information to be sent into a low-dimensional semantic coding signal to be sent after the signal transmitting end obtains the semantic information to be sent, and then transmits the semantic coding signal to be sent to the signal receiving end through the main channel. For example, if the semantic information to be sent is x, the semantic information to be sent is Semantic coding signal to be sent The bandwidth compression ratio of the first encoder is defined as K / N; wherein K and N represent the dimensions of the semantic information to be sent and the semantic coded signal to be sent, respectively. Then, the process of the first encoder compressing the semantic information to be sent x into the semantic coded signal to be sent z can be expressed as:

[0062]

[0063] Wherein, z is the semantic coding signal to be sent; represents the first encoder; θ1 represents the encoder parameter of the first encoder; x is the semantic information to be sent.

[0064] In the present invention, the cooperative jammer is intended to cooperatively interfere with the transmission of the semantic information to be sent, and acts as a cooperative node. When the signal sending end obtains the semantic information to be sent, the cooperative jammer synchronously generates a counter-interference noise.

[0065] In a preferred embodiment, the cooperative jammer includes a second encoder;

[0066] Generating anti-interference noise when acquiring semantic information to be sent includes: acquiring random noise when acquiring semantic information to be sent; inputting the random noise into a second encoder so that the second encoder outputs the anti-interference noise.

[0067] Specifically, a second encoder is deployed in the cooperative jammer; when it is detected that the signal transmitter obtains the semantic information to be sent, the cooperative jammer randomly samples a random noise from the Gaussian distribution and inputs the random noise into the second encoder, so that the second encoder generates a counter-interference noise according to the input random noise. Exemplarily, if the random noise is s, Then the process of the second encoder generating anti-interference noise based on random noise can be expressed as:

[0068]

[0069] Among them, zj represents the resistance to interference noise, represents the second encoder; s is random noise; φ represents the encoder parameter of the second encoder.

[0070] It should be noted that the Gaussian signal is selected to sample the random noise because the Gaussian signal has good statistical characteristics and can ensure that the generated noise is random and unpredictable, thereby improving the interference effect.

[0071] While the signal transmitting end sends the semantically coded signal to be sent, the cooperative jammer synchronously transmits the generated anti-interference noise to the signal receiving end through the main channel.

[0072] In a preferred embodiment, the semantic information to be sent and the anti-interference noise are sent to the signal receiving end at the same frequency.

[0073] Specifically, by sending the semantic information to be sent and the anti-interference noise to the signal receiving end at the same frequency, the anti-interference noise can more fully interfere with the semantic coding signal to be sent, thereby achieving a better collaborative interference effect.

[0074] If the signal transmitter transmits at a specified power P s The signal receiving end sends the semantically coded signal z to be sent, and the cooperative jammer sends power P j Send anti-interference noise to the signal receiving end, the noise power on the main channel is Then, the signal-to-noise ratio result can be obtained to evaluate the quality of the signal at the signal receiving end and the performance of the communication link; wherein, the calculation formula of the signal-to-noise ratio is:

[0075]

[0076] Among them, SINR stands for signal-to-noise ratio. A higher SINR means better quality of the received signal and lower bit error rate, which is conducive to correct decoding of the signal and recovery of the original information. Therefore, SINR should be increased as much as possible while ensuring the interference effect.

[0077] When the signal receiving end receives a mixed signal consisting of a semantically coded signal to be sent and anti-interference noise, it extracts the mixed signal to obtain semantic information to be received corresponding to the semantic information to be sent.

[0078] In a preferred embodiment, the signal receiving end includes a first decoder and a denoiser;

[0079] Extracting semantic information to be received corresponding to the semantic information to be sent according to the mixed signal, including: inputting the mixed signal into a denoiser so that the denoiser outputs a semantic coded signal to be received corresponding to the semantic coded signal to be sent; inputting the semantic coded signal to be received into a first decoder so that the first decoder outputs the semantic information to be received.

[0080] Specifically, a denoiser and a first decoder are deployed in the signal receiving end. When the signal receiving end receives a mixed signal, the mixed signal is first input into the denoiser so that the denoiser removes the anti-interference noise in the mixed signal to obtain a semantically coded signal to be received; the semantically coded signal to be received is then input into the first decoder so that the first decoder decodes the semantically coded signal to be received to obtain semantic information to be received.

[0081] It should be noted that, since additive Gaussian white noise exists during the transmission of the main channel, the mixed signal finally received at the receiving end can be essentially expressed as:

[0082]

[0083] in, represents a mixed signal; z is the semantic coding signal to be sent; z j To combat interference noise; b is the additive white Gaussian noise vector of the main channel, which obeys the complex Gaussian distribution represents the noise power on the main channel; I is the unit matrix.

[0084] Exemplarily, after receiving the mixed signal, the denoiser denoises the mixed signal and then combines with the first decoder to reconstruct the semantic information to be received, which can be expressed as:

[0085]

[0086] in, Indicates the semantic information to be received, represents a first decoder; represents a denoiser; represents the mixed signal; μ is the parameter of the denoiser; θ2 is the decoder parameter of the first decoder.

[0087] In a preferred embodiment, the semantic communication system further includes: a wiretap simulator; the wiretap simulator is used to obtain the mixed signal and output a wiretap simulation signal according to the mixed signal.

[0088] In a preferred embodiment, the wiretap simulator includes a second decoder; and outputting the wiretap simulation signal according to the mixed signal includes: inputting the mixed signal into the second decoder so that the second decoder outputs the wiretap simulation signal.

[0089] Specifically, the semantic communication system based on cooperative interference of the present invention also includes a wiretap simulator, which is used to form a wiretap channel between the wiretap simulator and the signal sending end, and is used to simulate stealing the semantic information to be sent from the wiretap channel. The wiretap simulator includes a second decoder. After acquiring a mixed signal containing the semantic information to be sent and the anti-interference noise, the wiretap simulator processes the mixed signal through the second decoder to obtain a wiretap simulation signal.

[0090] Exemplarily, the mixed signal finally obtained by the eavesdropping simulator can be expressed as:

[0091]

[0092] in, represents the mixed signal obtained by the eavesdropping simulator; z is the semantic coding signal to be sent; z j To combat interference noise; e is the additive white Gaussian noise vector of the eavesdropping channel, which obeys the complex Gaussian distribution is the noise power on the eavesdropping channel; I is the unit matrix.

[0093] Then the process of the eavesdropping simulator reconstructing the eavesdropping simulation signal based on the obtained mixed signal can be expressed as:

[0094]

[0095] in, Indicates eavesdropping on analog signals; represents a second decoder that uses all the information of the first decoder, including its parameter set θ2, so that it has the same decoding capability as the first decoder; Represents the mixed signal obtained by the wiretap simulator.

[0096] It can be seen that since the eavesdropping simulator does not include a denoiser, the eavesdropping simulation signal it finally outputs carries noise. The collaborative interference effect can be determined by the eavesdropping simulation signal, the semantic information to be received and the semantic information to be sent.

[0097] In order to make the semantic information to be received obtained by the signal receiving end as equivalent as possible to the semantic information to be sent, and the interference counter-noise sent by the cooperative jammer can have the greatest interference effect on the eavesdropper, the construction process of the first encoder, the first decoder, the second encoder and the denoiser is particularly important. This process is described in detail below.

[0098] In a preferred embodiment, the construction of the first encoder and the first decoder includes: obtaining a semantic information training sample set; wherein the semantic information training sample set includes a plurality of semantic information samples; constructing an initial first encoder and an initial first decoder; using the semantic information samples as input to the initial first encoder so that the initial first encoder outputs semantic coding signal samples, using the semantic coding signal samples as input to the initial first decoder so that the initial first decoder outputs semantic information prediction samples corresponding to the semantic information samples, and jointly training the initial first encoder and the initial first decoder until the first joint loss function of the initial first encoder and the initial first decoder is minimized, thereby generating a first encoder and a first decoder; wherein, in each joint training process, the loss of the current joint training is determined based on the semantic information samples and the semantic information prediction samples corresponding to the semantic information samples.

[0099] Specifically, Figure 2 As shown, it is a schematic diagram of the training process of the first encoder and the first decoder of the present invention. First, a semantic information training sample set is obtained, and the semantic information training sample set includes a large number of semantic information samples. Preferably, when the semantic information is image-type semantic information, the corresponding semantic information training sample set can use CelebAMask-HQ or CIFAR-10. Among them, CelebAMask-HQ is a large-scale face image dataset, and CIFAR-10 is a color image dataset of universal objects.

[0100] Secondly, an initial first encoder and an initial first decoder are constructed. The initial first encoder is composed of a convolutional neural network, which is used to extract high-level semantic features of input data and compress high-dimensional input data into low-dimensional data output. The initial first decoder is also composed of a convolutional neural network, which is used to reconstruct the output data of the initial second encoder into the input data of the initial first encoder.

[0101] In the joint training process of the initial first encoder and the initial first decoder, for each iterative training, a semantic information sample x′ is obtained from the semantic information training sample set, and the semantic information sample x′ is input into the initial first encoder, so that the initial first encoder outputs a semantic coding signal sample z′ from the semantic information sample x′; the semantic coding signal sample z′ is input into the initial first decoder, so that the initial first decoder reconstructs a semantic information prediction sample corresponding to the semantic information sample x′ according to the semantic coding signal sample z′ Predict samples based on semantic information The error between the initial first encoder and the semantic information sample x′ can determine the loss of the current joint training. The encoder parameters of the initial first encoder and the decoder parameters of the first decoder are adjusted according to the loss of the current joint training before the next round of training is carried out until the first joint loss function of the initial first encoder and the initial first decoder is minimized. The training of the initial first encoder and the initial first decoder is completed, and the initial first encoder at this time is used as the first encoder in the signal sending end, and the initial first decoder at this time is used as the first decoder in the signal receiving end. The encoder parameters of the first encoder are θ1 and the decoder parameters of the first decoder are θ2.

[0102] Preferably, the first joint loss function is specifically:

[0103]

[0104] in, represents the first joint loss function; λ is a trade-off coefficient; represents the cross entropy loss function; represents the mean square error loss function; is used to measure the semantic information sample x′ and the semantic information prediction sample function of the difference between .

[0105] In a preferred embodiment, the construction of the second encoder and the denoiser includes: freezing the encoder parameters of the first encoder and the decoder parameters of the second decoder; obtaining a semantic information training sample set and a random noise sample; constructing an initial second encoder and an initial denoiser; using the random noise sample as the input of the initial second encoder, using the adversarial interference noise prediction sample corresponding to the random noise sample as the output of the initial second encoder, using the mixed signal sample constructed by the semantic information sample and the adversarial interference noise prediction sample as the input of the initial denoiser, using the semantic coding prediction signal corresponding to the semantic information sample as the output, using the mixed signal sample as the input of the second decoder, using the eavesdropping simulation prediction signal of the second decoder as the output, and performing joint adversarial training on the initial second encoder and the initial denoiser until the second joint loss function of the initial second encoder and the initial denoiser is minimized, thereby generating the second encoder and denoiser.

[0106] Specifically, first, the decoder parameter θ2 of the first decoder is assigned to the second decoder, and the encoder parameter θ1 of the first encoder and the decoder parameter θ2 of the second decoder are frozen. Then, a semantic information training sample set and a random noise sample s′ are obtained to construct an initial second encoder and an initial denoiser. Figure 3 As shown, when the second encoder and the denoiser are jointly trained against each other, for each training session, the first encoder obtains a semantic information sample x′, inputs the semantic information sample x′ into the first encoder for encoding, and generates a coded signal z″. The collaborative jammer obtains a random noise sample s′ from a Gaussian distribution, and inputs the random noise sample s′ into the initial second encoder, so that the initial second encoder generates an adversarial interference noise prediction sample z j ′; Combine the coded signal z″ and the anti-interference noise prediction sample z j ′ is transmitted to the initial denoiser through the main channel. At this time, the mixed signal sample can be expressed as follows based on the additive Gaussian white noise vector of the main channel:

[0107]

[0108] in, represents the mixed signal sample; z″ is the coded signal; z j ′ is the prediction sample against interference noise; κ b is the additive white Gaussian noise vector of the main channel, which obeys the complex Gaussian distribution represents the noise power on the main channel; I is the unit matrix.

[0109] In mixed signal samples After inputting the initial denoiser, the initial denoiser removes as many mixed signal samples as possible. κb and z j ′, get the semantic coding prediction signal After that, the semantic encoding prediction signal The first decoder is input into the first decoder so that the first decoder outputs semantic information corresponding to the semantic coding prediction signal according to the semantic coding prediction signal. The mixed signal samples are used as the input of the second decoder, and the wiretap simulation prediction signal of the second decoder is used as the It should be noted that, since the mixed signal sample input to the second decoder contains the additive Gaussian white noise vector κ of the eavesdropping channel e , therefore, the mixed signal sample input to the second decoder should be expressed as:

[0110]

[0111] in, represents the mixed signal sample input to the second decoder; z″ is the encoded signal; z j ′ is the prediction sample against interference noise; κ e is the additive white Gaussian noise vector of the eavesdropping channel, which obeys the complex Gaussian distribution is the noise power on the eavesdropping channel; I is the unit matrix.

[0112] In this training method, during each joint adversarial training, adversarial interference noise prediction samples z are added j ′The denoising ability of the denoiser and the ability of the second decoder to decipher the data are observed.

[0113] In a preferred embodiment, when the initial second encoder and the initial denoiser are jointly adversarially trained, the encoder parameters of the initial second encoder and the denoising parameters of the initial denoiser are adjusted according to the eavesdropping simulation prediction signal, the semantic coding prediction signal and the semantic information sample.

[0114] Specifically, when the initial second encoder and the initial denoiser are jointly trained against each other, the eavesdropping simulation prediction signal The similarity between the semantic information sample x′ and the initial second encoder is used to adjust the encoder parameters of the second encoder to strengthen the interference strength and reduce the amount of stolen information; according to the semantic coding prediction signal The similarity between the semantic information sample x′ and the initial denoiser is used to adjust the denoising parameters of the initial denoiser to enhance the denoising ability of the initial denoiser.

[0115] Through continuous joint adversarial training, until the second joint loss function of the initial second encoder and the initial denoiser is minimized, the training of the initial second encoder and the initial denoiser is completed, and the initial second encoder at this time is used as the second encoder of the collaborative jammer, and the denoiser at this time is used as the denoiser in the receiving end after absorption.

[0116] Preferably, the second joint loss function is determined based on the loss function of the second decoder and the loss function of the denoiser.

[0117] The loss function of the second decoder is specifically:

[0118]

[0119] in, represents the loss function of the second decoder; λ is a trade-off coefficient; represents the cross entropy loss function; represents the mean square error loss function; v represents the error label, which is used to measure the interference effect on the second decoder; represents the eavesdropping simulation prediction signal; Used to measure the mean square error between the eavesdropping simulation prediction signal and matrix 1.

[0120] The loss function of the denoiser is:

[0121]

[0122] in, represents the loss function of the denoiser; λ is a trade-off coefficient; represents the cross entropy loss function; represents the mean square error loss function; is used to measure the semantic information sample x′ and the semantic coding prediction signal function of the difference between .

[0123] By implementing the present invention, anti-interference noise can be added when the semantic information to be sent is sent in a collaborative interference and physical layer security manner, so that the semantic information to be sent and the anti-interference noise exist in the main channel at the same time. The anti-interference noise confuses the eavesdropper and makes it difficult for the eavesdropper to directly obtain the semantic information to be sent, thereby improving the security and transmission reliability of the semantic information transmitted through the wireless channel.

[0124] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0125] Those skilled in the art can clearly understand that for the sake of convenience and brevity, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0126] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0127] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.

[0128] The memory can be used to store the computer program, and the processor realizes various functions of the terminal device by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0129] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by the processor, the steps of each of the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium.

[0130] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A semantic communication system based on cooperative interference, characterized in that: include: Signal transmitter, signal receiver and cooperative jammer; The signal transmitting end is used to obtain the semantic information to be sent, and send the semantic information to be sent to the signal receiving end through the main channel; wherein the main channel is a wireless channel between the signal transmitting end and the signal receiving end; The cooperative jammer is used to generate anti-interference noise when acquiring semantic information to be sent, and send the anti-interference noise to the signal receiving end through the main channel when the signal sending end sends the semantic information to be sent to the signal receiving end through the main channel; The signal receiving end is used to receive a mixed signal including the semantic information to be sent and the anti-interference noise, and extract the semantic information to be received corresponding to the semantic information to be sent according to the mixed signal.

2. A semantic communication system based on collaborative interference as claimed in claim 1, characterized in that: The signal transmitting end comprises a first encoder; The acquiring the semantic information to be sent and sending the semantic information to be sent to the signal receiving end through the main channel includes: Obtaining semantic information to be sent; Inputting the semantic information to be sent into a first encoder so that the first encoder outputs a semantic coding signal to be sent; The semantically coded signal to be sent is sent to a signal receiving end through a main channel.

3. A semantic communication system based on collaborative interference as claimed in claim 2, characterized in that: The cooperative jammer includes a second encoder; The generating of anti-interference noise when acquiring the semantic information to be sent includes: Acquire random noise when acquiring semantic information to be sent; The random noise is input into a second encoder so that the output of the second encoder is resistant to interference noise.

4. A semantic communication system based on cooperative interference as claimed in claim 3, characterized in that: The signal receiving end includes a first decoder and a denoiser; The extracting the semantic information to be received corresponding to the semantic information to be sent according to the mixed signal includes: Inputting the mixed signal into a denoiser so that the denoiser outputs a semantically coded signal to be received corresponding to the semantically coded signal to be sent; The semantically coded signal to be received is input into a first decoder, so that the first decoder outputs semantic information to be received.

5. A semantic communication system based on cooperative interference as claimed in claim 4, characterized in that: The construction of the first encoder and the first decoder includes: Acquire a semantic information training sample set; wherein the semantic information training sample set includes a plurality of semantic information samples; constructing an initial first encoder and an initial first decoder; The semantic information sample is used as the input of the initial first encoder so that the initial first encoder outputs a semantic coding signal sample, and the semantic coding signal sample is used as the input of the initial first decoder so that the initial first decoder outputs a semantic information prediction sample corresponding to the semantic information sample. The initial first encoder and the initial first decoder are jointly trained until the first joint loss function of the initial first encoder and the initial first decoder is minimized, thereby generating a first encoder and a first decoder; wherein, in each joint training process, the loss of the current joint training is determined according to the semantic information sample and the semantic information prediction sample corresponding to the semantic information sample.

6. A semantic communication system based on cooperative interference as claimed in claim 5, characterized in that: The semantic communication system further includes: a wiretapping simulator; The wiretapping simulator is used to obtain the mixed signal and output a wiretapping simulation signal according to the mixed signal.

7. A semantic communication system based on collaborative interference as claimed in claim 6, characterized in that: The wiretapping simulator includes a second decoder; The method outputs a wiretap simulation signal according to the mixed signal, comprising: The mixed signal is input into a second decoder so that the second decoder outputs a wiretap simulation signal.

8. A semantic communication system based on collaborative interference as claimed in claim 7, characterized in that: The construction of the second encoder and the denoiser includes: freezing encoder parameters of the first encoder and decoder parameters of the second decoder; Obtain semantic information training sample sets and random noise samples; Constructing an initial second encoder and an initial denoiser; The random noise samples are used as the input of the initial second encoder, the adversarial interference noise prediction samples corresponding to the random noise samples are used as the output of the initial second encoder, the mixed signal samples constructed by the semantic information samples and the adversarial interference noise prediction samples are used as the input of the initial denoiser, the semantic coding prediction signal corresponding to the semantic information samples is used as the output, the mixed signal samples are used as the input of the second decoder, and the eavesdropping simulation prediction signal of the second decoder is used as the output. The initial second encoder and the initial denoiser are jointly adversarially trained until the second joint loss function of the initial second encoder and the initial denoiser is minimized, thereby generating the second encoder and denoiser.

9. A semantic communication system based on cooperative interference as claimed in claim 8, characterized in that: When the initial second encoder and the initial denoiser are jointly trained against each other, encoder parameters of the initial second encoder and denoising parameters of the initial denoiser are adjusted according to the eavesdropping simulation prediction signal, the semantic coding prediction signal and the semantic information sample.

10. The semantic communication system based on cooperative interference according to claim 1, characterized in that: The semantic information to be sent and the anti-interference noise are sent to the signal receiving end at the same frequency.

Images