Safety interruption probability prediction method and system, and electronic equipment

Abstract

The present invention relates to the field of communications, and provides a method and system for predicting the probability of a safe interruption. The method comprises: creating a communication data set based on collected target data, wherein the communication data set comprises a training set; training the training set based on a deep separable convolution and a dilated convolution, and then obtaining a prediction model by applying a Transformer model; and predicting the probability of a safe interruption based on the prediction model. The method solves the defect that it is difficult to achieve real-time and accurate prediction of the security performance of the Internet of Things in the prior art, and by predicting the probability of a safe interruption, the security performance of the Internet of Things can be accurately predicted.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a method and system for predicting the probability of safety interruption, and electronic equipment. Background Art

[0002] With the full integration and application of cutting-edge technologies such as big data, Internet+, and blockchain, the mobile Internet of Things (IoT) is increasingly showing a strong influence and vitality. However, due to the openness and diversity of the mobile Internet of Things itself, it may cause serious information security issues.

[0003] However, the complexity of IoT mobile communication environment makes it difficult to use deep learning algorithms for security performance analysis and prediction. At present, the research on communication security performance prediction algorithms based on physical layer security is still in its infancy. Especially for complex and changeable IoT communication scenarios, how to achieve real-time and accurate IoT security performance prediction has not been well solved. Summary of the invention

[0004] The present invention provides a method and system for predicting the probability of security interruption, and an electronic device, so as to solve the defect that it is difficult to realize real-time and accurate prediction of the security performance of the Internet of Things in the prior art. By predicting the probability of security interruption, the security performance of the Internet of Things can be accurately predicted.

[0005] The safety interruption probability prediction method provided by the present invention includes:

[0006] Creating a communication data set based on the collected target data, the communication data set including a training set;

[0007] After training the training set based on depthwise separable convolution and dilated convolution, a prediction model is obtained by applying a Transformer model;

[0008] Predict safety interruption probability based on prediction model.

[0009] According to the safety interruption probability prediction method provided by the present invention, after training the training set based on depthwise separable convolution and dilated convolution, a prediction model is obtained by applying a Transformer model, including:

[0010] Performing depth-wise separable convolution on the training set to determine first data;

[0011] Perform a dilated convolution on the training set to determine the second data;

[0012] The first data and the second data are encoded and decoded based on the Transformer model, and training is performed based on the decoding result.

[0013] According to the safety interruption probability prediction method provided by the present invention, a depth-separable convolution is performed on a training set to determine first data, including:

[0014] Perform A depth-wise separable convolutions on the training set, where A is a positive integer;

[0015] After batch normalization, the first data is obtained by activation through an activation function.

[0016] According to the method for predicting the probability of safety interruption provided by the present invention, performing a hole convolution on the communication data set to determine the second data includes:

[0017] Perform A dilated convolutions on the training set, where A is a positive integer;

[0018] After batch normalization, the second data is activated by the activation function.

[0019] According to the safety interruption probability prediction method provided by the present invention, encoding and decoding the first data and the second data based on the Transformer model include:

[0020] Perform feature fusion on the first data and the second data to obtain a fusion matrix;

[0021] The fusion matrix is ​​encoded and decoded through the Transformer model.

[0022] According to the safety interruption probability prediction method provided by the present invention, target data is collected in the following manner:

[0023] Establish an IoT secure communication system model, which includes the source, relay node, destination, and eavesdropping end;

[0024] Determine the signal-to-noise ratio of the IoT secure communication system model and calculate the probability density function and cumulative distribution function of the signal-to-noise ratio;

[0025] The security interruption probability of the IoT security communication system model is determined, and the target data is determined based on the security interruption probability, the probability density function and the cumulative distribution function of the IoT security communication system model.

[0026] According to the safety interruption probability prediction method provided by the present invention, the communication data set also includes a test set, the training set is trained based on the depth separable convolution and the hole convolution, and then further includes:

[0027] Use the test set to evaluate the performance of the trained prediction model. If the prediction model does not meet the preset performance indicators, retrain it.

[0028] According to the method for predicting the probability of safety interruption provided by the present invention, a communication data set is created based on the collected target data, and then the method further includes:

[0029] Convert the communication data set from two-dimensional data to three-dimensional data.

[0030] The present invention also provides a safety interruption probability prediction system, comprising:

[0031] A creation module, used to create a communication data set based on the collected target data, the communication data set includes a training set and a test set;

[0032] The training module is used to train the training set based on depthwise separable convolution and dilated convolution, and then apply the Transformer model to obtain the prediction model;

[0033] The prediction module is used to predict the probability of safety interruption based on the prediction model.

[0034] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, any of the above-mentioned security interruption probability predictions is implemented.

[0035] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which implements any of the above-mentioned safety interruption probability predictions when executed by a processor.

[0036] The present invention also provides a computer program product, comprising a computer program, which implements any of the above-mentioned security interruption probability predictions when executed by a processor.

[0037] The safety interruption probability prediction method provided by the present invention can create a communication data set based on the collected target data. Then, when training the communication data set, deep separable convolution and dilated convolution can be applied. Through the combination of the two technologies, on the one hand, the accuracy of the prediction model can be improved, and the accuracy of subsequent safety interruption probability prediction can be improved. On the other hand, the number of model parameters can be effectively reduced, and the amount of calculation for model training and subsequent safety interruption probability prediction can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1This is one of the flow charts of the safety interruption probability prediction method provided by the present invention;

[0040] Figure 2 This is the second flow chart of the safety interruption probability prediction method provided by the present invention;

[0041] Figure 3 It is one of the network structure diagrams of the prediction model provided by the present invention;

[0042] Figure 4 This is the second schematic diagram of the network structure of the prediction model provided by the present invention;

[0043] Figure 5 This is the third schematic diagram of the network structure of the prediction model provided by the present invention;

[0044] Figure 6 This is the fourth schematic diagram of the network structure of the prediction model provided by the present invention;

[0045] Figure 7 It is a structural schematic diagram of the IoT secure communication system model provided by the present invention;

[0046] Figure 8 This is the fifth schematic diagram of the network structure of the prediction model provided by the present invention;

[0047] Fig. 9 It is a structural schematic diagram of the safety interruption probability prediction system provided by the present invention;

[0048] Fig.10 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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.

[0050] Figure 1 This is one of the flow charts of the safety interruption probability prediction method provided by the present invention.

[0051] Figure 2 This is the second flow chart of the safety interruption probability prediction method provided by the present invention.

[0052] like Figure 1 and Figure 2 As shown, this embodiment provides a method for predicting the probability of a safety interruption, including:

[0053] Step 101, creating a communication data set based on the collected target data, where the communication data set includes a training set;

[0054] Step 102, after training the training set based on depthwise separable convolution and dilated convolution, applying the Transformer model to obtain a prediction model;

[0055] Step 103: predict the safety interruption probability based on the prediction model.

[0056] The depthwise separable convolution used in this embodiment is an algorithm obtained by improving the standard convolution calculation in the convolutional neural network.

[0057] In implementation, depthwise separable convolution can include two processes, namely depthwise convolution and pointwise convolution. By applying depthwise separable convolution, feature fusion between channels can be achieved, the signal direction dimension can be changed, and the calculation complexity can be effectively reduced. Compared with conventional convolution, the calculation amount of depthwise separable convolution only uses about one-third of the calculation amount to complete the same training task.

[0058] Dilated Convolution can also be called expanded convolution or dilated convolution. Compared with conventional convolution, the convolution kernel elements of conventional convolution are closely connected, while dilated convolution can expand the convolution kernel by adding some spaces between dense convolution kernel elements, or adding some digital zeros. In this way, the receptive field can be increased while keeping the size of the feature map unchanged. The receptive field refers to the area range of a pixel on the feature map corresponding to the input space, that is, the mapping of feature map pixels to the input area. By applying dilated convolution, the downsampling and upsampling operations in the conventional convolution process can be replaced, and the information loss problem that may exist in the kernel convolution-pooling-upsampling process in the conventional convolution process can be solved.

[0059] In this embodiment, after creating a communication data set based on the collected target data, training is not performed directly based on the communication data set. Instead, the communication data set can be first divided into a training set and a test set. The training set can then be trained based on depthwise separable convolution and dilated convolution. After the training is completed and the prediction model is obtained, the effect of the model can be tested through the test set. In this way, the prediction model obtained through training can be more accurate.

[0060] In this embodiment, a communication data set can be created based on the collected target data. Then, when training the communication data set, deep separable convolution and dilated convolution can be applied. By combining the two technologies, on the one hand, the accuracy of the prediction model can be improved, and the accuracy of the subsequent safety interruption probability prediction can be improved. On the other hand, the number of model parameters can be effectively reduced, and the amount of calculation for model training and subsequent safety interruption probability prediction can be reduced.

[0061] In an exemplary embodiment, after training the training set based on depthwise separable convolution and dilated convolution, a prediction model is obtained by applying a Transformer model, including:

[0062] Performing depth-wise separable convolution on the training set to determine first data;

[0063] Performing a dilated convolution on the communication data set to determine second data;

[0064] The first data and the second data are encoded and decoded based on the Transformer model, and training is performed based on the decoding result.

[0065] Figure 3 It is one of the network structure diagrams of the prediction model provided by the present invention.

[0066] Figure 4 This is the second schematic diagram of the network structure of the prediction model provided by the present invention.

[0067] like Figure 3As shown, the prediction model provided in this embodiment is a dual-branch network composed of three network structures, wherein the three network structures are a deep separable convolution module, a hole convolution module and a Transformer module, wherein the deep separable convolution module can replace the standard convolution process. Compared with the standard convolution, the deep separable convolution provided in this embodiment can effectively reduce the computational complexity and improve the performance of the convolution. The hole convolution module can reduce the number of parameters and can be used to solve the information loss problem in the ordinary small convolution kernel convolution-pooling-upsampling. It can increase the receptive field while keeping the size of the feature map unchanged, thereby replacing the downsampling and upsampling operations. In the implementation, the deep separable convolution module and the hole convolution module constitute the dual-branch structure of the prediction model, so that the model can learn from features of different scales. The deep separable convolution module is mainly responsible for capturing local features, while the hole convolution module is responsible for capturing global features of larger scales. This multi-scale feature learning can improve the generalization ability and prediction performance of the model. At the same time, since the depthwise separable convolution module and the atrous convolution module can reduce the amount of calculation and the number of parameters without sacrificing performance, the dual-branch structure can more effectively utilize the model parameters. In addition, the dual-branch structure is highly flexible and can appropriately adjust the depthwise separable convolution module and the atrous convolution module according to task requirements and computing resources. For example, parameters such as the convolution kernel size and dilation rate can be adjusted to optimize the performance of the model.

[0068] The main structure of the Transformer module is the Transformer model, such as Figure 4 As shown in the figure, the Transformer model includes an encoder and a decoder. In the Transformer model, the features of the input sequence can be captured through the self-attention mechanism. Compared with the traditional recurrent neural network model, the Transformer model has more powerful model expression capabilities and more efficient parallel computing capabilities.

[0069] In an exemplary embodiment, the first data obtained by the depthwise separable convolution module after performing depthwise separable convolution on the training set can be in matrix form, and the second data obtained by the dilated convolution module after performing dilated convolution on the training set can also be in matrix form. After the two branches are calculated, the first data and the second data can be concatenated in the Transformer module to complete feature fusion and obtain a fusion matrix, where the fusion matrix can be a 400×64×4×4 matrix, and then the fusion matrix is ​​encoded and decoded based on the Transformer model. The encoding dimension of the Transformer model can be set to 512, and the head can be set to 2.

[0070] In implementation, the Transformer module may also include a batch normalization layer, a ReLU activation layer, and a fully connected layer. In the batch normalization layer, the fusion matrix may be standardized, which may reduce the complexity of training and speed up training. After that, the weighted input from the node may be converted into the activation of the node or output of the input through the ReLU activation layer through the ReLU (rectified linear unit) function. The ReLU function is a piecewise linear function. If the input is positive, it will directly output, otherwise, it will output zero.

[0071] In implementation, before encoding and decoding the fusion matrix, the fusion matrix can be transformed to readjust the number of rows, columns and dimensions of the matrix. For example, the fusion matrix can be transformed through the reshape function in MATLAB. The input data of the network model in this project is composed of relevant factors that affect the probability of safe interruption of the channel, such as power allocation coefficient, signal-to-noise ratio, position gain coefficient and safety capacity threshold. In general, the strip input matrix composed of these influencing factors can be directly input for training. However, since the model used in this article is more suitable for input data in the form of pictures, and the picture form has richer information features than strip data, the correlation between data can be increased in the convolutional network, which is conducive to extracting more effective features. For example, the fusion matrix can be converted from a 400×64×4×4 matrix to a 1×400×512 matrix.

[0072] In an exemplary embodiment, performing a depthwise separable convolution on a training set to determine first data includes:

[0073] Perform A depth-wise separable convolutions on the training set, where A is a positive integer;

[0074] After batch normalization, the first data is obtained by activation through an activation function.

[0075] Figure 5 This is the third network structure diagram of the prediction model provided by the present invention.

[0076] like Figure 5 As shown, in this embodiment, in the process of performing depth-wise separable convolution on the training set in the depth-wise separable convolution module, A depth-wise separable convolutions can be performed, A can be a positive integer. Exemplarily, the value of A can be 3. In other embodiments, the value of A can also be customized according to actual needs. By performing multiple depth-wise separable convolutions, the training effect can be further enhanced.

[0077] In an exemplary embodiment, Figure 3As shown in the figure, the depthwise separable convolution module also includes a batch normalization layer and a ReLU activation layer. After the input data undergoes three depthwise separable convolutions, it is normalized by the batch normalization layer and activated by the ReLU activation function in the ReLU activation layer. After activation, a 400×64×2×2 matrix can be obtained.

[0078] In an exemplary embodiment, performing a dilated convolution on a communication data set to determine second data includes:

[0079] Perform A dilated convolutions on the training set, where A is a positive integer;

[0080] After batch normalization, the second data is activated by the activation function.

[0081] Figure 6 This is the fourth network structure diagram of the prediction model provided by the present invention.

[0082] like Figure 6 As shown, in this embodiment, in the process of performing dilated convolution on the training set in the dilated convolution module, A dilated convolutions can be performed, A can be a positive integer. For example, the value of A can be 3. In other embodiments, the value of A can also be customized according to actual needs. By performing multiple dilated convolutions, the training effect can be further enhanced.

[0083] In implementation, the dilation rates of the three dilated convolutions can be 1, 2, and 2, respectively.

[0084] In an exemplary embodiment, Figure 3 As shown in the figure, the dilated convolution module also includes a batch normalization layer and a ReLU activation layer. After the input data undergoes three dilated convolutions, it is normalized by the batch normalization layer and activated by the ReLU activation function in the ReLU activation layer. After activation, a 400×64×2×2 matrix can be obtained.

[0085] In an exemplary embodiment, the communication data set also includes a test set, and training a prediction model based on the communication data set includes:

[0086] Use the test set to evaluate the performance of the trained prediction model. If the prediction model does not meet the preset performance indicators, retrain it.

[0087] In the above embodiments, it has been disclosed that the communication data set includes a training set and a test set, and a prediction model can be obtained by training based on the training set. In implementation, the test set can also be used to test the trained prediction model, that is, the data of the test set can be predicted based on the prediction model, and then the true value and the predicted value can be compared according to multiple performance indicators to judge the prediction performance of the prediction model. After the comparison, if it is determined that the prediction model does not meet the performance indicators, the prediction model can be further trained and then tested again until the prediction performance of the prediction model meets the requirements.

[0088] Exemplarily, the performance indicators provided in this embodiment may include mean square error, determination coefficient, absolute error and execution time, wherein the calculation formula of the mean square error satisfies the following formula (1):

[0089]

[0090] The calculation formula of the determination coefficient satisfies the following formula (2):

[0091]

[0092] Where k represents the number of samples in the test set, x n represents the true value of the samples in the test set,

[0093] Represents the predicted value obtained according to the prediction model.

[0094] In an exemplary embodiment, target data is collected by:

[0095] Establish an IoT secure communication system model, which includes the source, relay node, destination, and eavesdropping end;

[0096] Determine the signal-to-noise ratio of the IoT secure communication system model and calculate the probability density function and cumulative distribution function of the signal-to-noise ratio;

[0097] The security interruption probability of the IoT security communication system model is determined, and the target data is determined based on the security interruption probability, the probability density function and the cumulative distribution function of the IoT security communication system model.

[0098] Figure 7 It is a structural schematic diagram of the IoT security communication system model provided by the present invention.

[0099] like Figure 7 As shown, in this embodiment, an IoT secure communication system model is pre-established, and the model includes a source MS, a relay node MR, a destination MD, and an eavesdropping segment ME. The channel fading coefficient h of the model is (SR,RD,RE)，Obeys N-Nakagami distribution.

[0100] In the implementation, in the IoT secure communication system model, the source MS can communicate with the relay node MR, and the communication process between the source MS and the relay node MR conforms to formula (3):

[0101]

[0102] Where x is the signal value transmitted by the source MS, K and E represent the power allocation coefficient and the system transmission power respectively, and n SR represents Gaussian noise with mean and expectation of 0 and 1 respectively, G SR represents the location gain from the source MS to the relay node MR, y SRi Represents the signal received by the relay node MR.

[0103] After receiving the signal sent by the source MS, the relay node MR can forward the signal to the destination MD by means of decode-and-forward (DF). The communication process between the relay node MR and the destination MD conforms to formula (4):

[0104]

[0105] The communication process between the relay node MR and the eavesdropping segment ME conforms to formula (5):

[0106]

[0107] where y RDi is the signal value received by the destination MD, y REi is the signal value received by the eavesdropping end ME, n RDi and n REi is a Gaussian noise with mean and expectation of 0 and 1 respectively, G RD represents the position gain from the relay node MR to the destination MD, G RE It represents the position gain from the relay node MR to the eavesdropping end ME. If the relay MR can demodulate the data correctly, ɑ=1; otherwise ɑ=0.

[0108] MS i The mutual information between MR and MR satisfies formula (6):

[0109]

[0110] In formula (6), γ SRi is the signal-to-noise ratio of the communication link between the source MS and the relay node MR. th , when the source MS i The mutual information I between the relay MRSRi >R th When the mutual information I SRi <R th When the relay MR cannot achieve correct decoding, an interruption occurs.

[0111] In implementation, the signal-to-noise ratio γ of the IoT security communication system model provided in this embodiment can also be k , k∈{D,E}, and then γ can be calculated according to the following formula (7) k The probability density function of γ is calculated according to the following formula (8): k Cumulative distribution function of :

[0112]

[0113]

[0114] m is the attenuation coefficient, N is the attenuation factor, Ω=E(|z| 2 ), z is a variable that conforms to the Nakagami distribution, E() represents the mean operation; G[·] represents Meijer's G function;

[0115] is the average signal-to-noise ratio of the MS-to-MR link, The calculation formula conforms to the following formula (9):

[0116]

[0117] is the average signal-to-noise ratio of the MR to Mk link, The calculation formula conforms to the following formula (10):

[0118]

[0119] in is the average signal-to-noise ratio, The calculation formula conforms to the following formula (11):

[0120]

[0121] E is the power and N0 is the power spectral density.

[0122] In addition, the instantaneous security capacity of the IoT security communication system model can also be calculated. The instantaneous security capacity of the IoT security communication system model provided in this embodiment satisfies the following formula (12):

[0123] C i(y D ,y E )=max{1n(1+γ Di )-1n(1+γ Ei ), 0} (12)

[0124] The safety outage probability (SOP) is a common method to evaluate the safety of communication systems. During the communication process, the instantaneous safety capacity is less than the safety outage threshold γ th The probability of is the safety interruption probability. In implementation, the calculation formula of SOP conforms to the following formula (13):

[0125]

[0126] Where β = exp(γ th ), exp() represents the exponential function with the natural constant e as the base.

[0127] In this embodiment, it can be determined that the lower bound closed expression of the security interruption probability of the IoT security communication system model conforms to the following formula (14):

[0128]

[0129] After that, the target data can be calculated based on the probability density function, the cumulative distribution function and the safety interruption probability. That is, by combining formula (7) and formula (8), the expression of the target data can be calculated to satisfy the following formula (15):

[0130]

[0131] Where W is the target data, and the calculation formula of V in the formula satisfies the following formula (16):

[0132]

[0133] In an exemplary embodiment, the communication data set created may be {T n}, T n =(X n ,y n ), where y n is the target parameter calculated in the above embodiment, X n is an input vector composed of multiple features, for example, X n The expression satisfies the following formula (17):

[0134] X n =(X n1 , X n2 , X n3 , X n4 , X n5 , Xn6 )(17)

[0135] Where X n1 Can be K, which is the power allocation coefficient, X n2 Can be G SR , that is, the position gain between the source MS and the relay node MR, X n3 Can be G RD , that is, the position gain between the relay node MR and the destination MD, X n4 Can be G RE , that is, the position gain between the relay node and the eavesdropping end RE, X n5 Can be r th , X n6 Can be That is, the average signal-to-noise ratio of the IoT secure communication system model.

[0136] Figure 8 This is the fifth network structure diagram of the prediction model provided by the present invention.

[0137] In an exemplary embodiment, Figure 8 As shown, a communication data set is created based on the collected target data, and then it also includes:

[0138] Convert the communication data set from two-dimensional data to three-dimensional data.

[0139] In implementation, the communication data set is two-dimensional data. In this embodiment, the two-dimensional data in the communication data set can be converted. After being converted into three-dimensional data, the three-dimensional data can contain more feature information than the two-dimensional data, and the correlation is usually higher, which is conducive to extracting more effective features. For example, the data in the communication data set can be converted from 1×6 two-dimensional data to 1×12×12 three-dimensional data.

[0140] The following is a description of the security interruption probability prediction system provided by the present invention. The security interruption probability prediction system described below and the security interruption probability prediction system described above can correspond to each other.

[0141] Fig. 9 It is a structural schematic diagram of the safety interruption probability prediction system provided by the present invention.

[0142] like Fig. 9 As shown, the safety interruption probability prediction system provided by this embodiment includes:

[0143] A creation module 901, used to create a communication data set based on the collected target data;

[0144] A training module 902 is used to train the training set based on depthwise separable convolution and dilated convolution, and then apply the Transformer model to obtain a prediction model;

[0145] The prediction module 903 is used to predict the safety interruption probability based on the prediction model.

[0146] In an exemplary embodiment, the creation module 901 is specifically used for:

[0147] Performing depth-wise separable convolution on the training set to determine first data;

[0148] Performing a dilated convolution on the communication data set to determine second data;

[0149] The first data and the second data are encoded and decoded based on the Transformer model, and training is performed based on the decoding result.

[0150] In an exemplary embodiment, the training module 902 is specifically used to:

[0151] Perform A depth-wise separable convolutions on the training set, where A is a positive integer;

[0152] After batch normalization, the first data is obtained by activation through an activation function.

[0153] In an exemplary embodiment, the training module 902 is specifically used to:

[0154] Perform A dilated convolutions on the training set, where A is a positive integer;

[0155] After batch normalization, the second data is activated by the activation function.

[0156] In an exemplary embodiment, the training module 902 is specifically used to:

[0157] Perform feature fusion on the first data and the second data to obtain a fusion matrix;

[0158] The fusion matrix is ​​encoded and decoded through the Transformer model.

[0159] In an exemplary embodiment, target data is collected by:

[0160] Establish an IoT secure communication system model, which includes the source, relay node, destination, and eavesdropping end;

[0161] Determine the signal-to-noise ratio of the IoT secure communication system model and calculate the probability density function and cumulative distribution function of the signal-to-noise ratio;

[0162] The security interruption probability of the IoT security communication system model is determined, and the target data is determined based on the security interruption probability, the probability density function and the cumulative distribution function of the IoT security communication system model.

[0163] In the exemplary embodiment, a performance evaluation module is also included, and the performance evaluation module is specifically used to:

[0164] Use the test set to evaluate the performance of the trained prediction model. If the prediction model does not meet the preset performance indicators, retrain it.

[0165] In an exemplary embodiment, a pre-processing module is also included, and the pre-processing module is specifically used for:

[0166] Convert the communication data set from two-dimensional data to three-dimensional data.

[0167] The specific implementation method of the security interruption probability prediction system provided in this embodiment can be implemented with reference to the above embodiment and will not be repeated here.

[0168] Fig.10 An example of a physical structure diagram of an electronic device is shown in FIG. Fig.10 As shown, the electronic device may include: a processor 1010, a communication interface 1020, a memory 1030 and a communication bus 1040, wherein the processor 1010, the communication interface 1020 and the memory 1030 communicate with each other through the communication bus 1040. The processor 1010 may call the logic instructions in the memory 1030 to execute the security interruption probability prediction method, which includes:

[0169] Creating a communication data set based on the collected target data, the communication data set includes a training set and a test set;

[0170] The training set is trained based on depthwise separable convolution and dilated convolution to obtain a prediction model;

[0171] Predict safety interruption probability based on prediction model.

[0172] In addition, the logic instructions in the above-mentioned memory 1030 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk.

[0173] On the other hand, the present invention further provides a computer program product, the computer program product includes a computer program, the computer program can be stored in a non-transitory computer-readable storage medium, when the computer program is executed by a processor, the computer can execute the security interruption probability prediction method provided by the above methods, the method includes:

[0174] Creating a communication data set based on the collected target data, the communication data set includes a training set and a test set;

[0175] The training set is trained based on depthwise separable convolution and dilated convolution to obtain a prediction model;

[0176] Predict safety interruption probability based on prediction model.

[0177] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for predicting the probability of security interruption provided by the above methods is implemented, and the method comprises:

[0178] Creating a communication data set based on the collected target data, the communication data set includes a training set and a test set;

[0179] The training set is trained based on depthwise separable convolution and dilated convolution to obtain a prediction model;

[0180] Predict safety interruption probability based on prediction model.

[0181] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., 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. Those of ordinary skill in the art may understand and implement it without creative effort.

[0182] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.

[0183] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting the probability of a safety interruption, characterized in that: include: Creating a communication data set based on the collected target data, wherein the communication data set includes a training set; After training the training set based on depthwise separable convolution and dilated convolution, a prediction model is obtained by applying a Transformer model; Predicting the probability of safety interruption based on the prediction model; The target data is collected in the following way: Establishing an IoT secure communication system model, wherein the IoT secure communication system model includes a signal source, a relay node, a destination end, and an eavesdropping end; Determine the signal-to-noise ratio of the IoT secure communication system model, and calculate the probability density function and cumulative distribution function of the signal-to-noise ratio; Determine the security interruption probability of the IoT security communication system model, and determine the target data based on the security interruption probability of the IoT security communication system model, the probability density function and the cumulative distribution function; The IoT secure communication system model includes a source, a relay node, a destination, and an eavesdropping end. The channel fading coefficient h of the model is (SR,RD,RE)， Obeys N-Nakagami distribution; The communication process between the source and the relay node conforms to formula (3): Among them, x is the signal value transmitted by the source, K and E represent the power allocation coefficient and the system's transmission power respectively, and n SR represents Gaussian noise with mean and expectation of 0 and 1 respectively, G SR represents the location gain from the source to the relay node, y SRi Indicates the signal received by the relay node; After the relay node receives the signal sent by the source, it forwards the signal to the destination by decoding and forwarding. The communication process between the relay node and the destination conforms to formula (4): The communication process between the relay node and the eavesdropping end conforms to formula (5): where y RDi is the signal value received by the destination, y REi is the signal value received by the eavesdropping end, n RDi and n REi is a Gaussian noise with mean and expectation of 0 and 1 respectively, G RD represents the position gain from the relay node to the destination, G RE It represents the position gain from the relay node to the eavesdropping end. If the relay node can demodulate the data correctly, ɑ=1; otherwise ɑ=0. MS i The mutual information between the relay nodes satisfies formula (6): In formula (6), γ SRi is the signal-to-noise ratio of the communication link between the source and the relay node. For a given threshold information rate R th , when the source MS i The mutual information between the relay nodes I SRi >R th When the mutual information I SRi <R th When the relay node cannot achieve correct decoding, an interruption occurs; The signal-to-noise ratio of the IoT security communication system model is collected, and the signal-to-noise ratio is γ k , k∈{D,E}, and γ is calculated according to the following formula (7): k The probability density function of γ is calculated according to the following formula (8): k Cumulative distribution function of : m is the attenuation coefficient, N is the attenuation factor, Ω=E(|z| 2 ), z is a variable that conforms to the Nakagami distribution, E() represents the mean operation; G[·] represents the Meijer's G function; is the average signal-to-noise ratio of the MS-to-MR link, The calculation formula conforms to the following formula (9): is the average signal-to-noise ratio of the link from the relay node to Mk, The calculation formula conforms to the following formula (10): in is the average signal-to-noise ratio, The calculation formula conforms to the following formula (11): E is power, N0 is power spectral density; The instantaneous security capacity of the IoT security communication system model satisfies the following formula (12): C i (c D ,c E )=max{ln(1+γ Di )-ln(1+γ Ei ),0} (12) The calculation formula of the safety interruption probability conforms to the following formula (13): Where β = exp(γ th ), exp() represents the exponential function with the natural constant e as the base; The closed-end expression of the lower bound of the security interruption probability of the IoT security communication system model conforms to the following formula (14): The expression of the target data calculated satisfies the following formula (15): Where W is the target data, and the calculation formula of V in the formula satisfies the following formula (16): After training the training set based on the depthwise separable convolution and the dilated convolution, applying the Transformer model to obtain the prediction model includes: Performing depth-wise separable convolution on the training set to determine first data; Performing a dilated convolution on the training set to determine second data; Encode and decode the first data and the second data based on the Transformer model, and perform training based on the decoding results; The encoding and decoding of the first data and the second data based on the Transformer model includes: Performing feature fusion on the first data and the second data to obtain a fusion matrix; The fusion matrix is ​​encoded and decoded through the Transformer model.

2. The method for predicting the probability of safety interruption according to claim 1, characterized in that: The performing a depth-wise separable convolution on the training set to determine the first data includes: Perform A depth-wise separable convolutions on the training set, where A is a positive integer; After batch normalization, the first data is obtained by activation through an activation function.

3. The method for predicting the probability of safety interruption according to claim 1, characterized in that: The performing a dilated convolution on the training set to determine the second data includes: Perform A dilated convolutions on the training set, where A is a positive integer; After batch normalization, the second data is obtained by activation through an activation function.

4. The method for predicting the probability of safety interruption according to claim 1, characterized in that: The communication data set also includes a test set, and the training set is trained based on depthwise separable convolution and dilated convolution, and then further includes: The test set is used to perform performance evaluation on the trained prediction model. If the prediction model does not meet the preset performance indicators, retraining is performed.

5. The method for predicting the probability of safety interruption according to claim 1, characterized in that: The step of creating a communication data set based on the collected target data further comprises: The communication data set is converted from two-dimensional data to three-dimensional data.

6. A safety interruption probability prediction system, applied to the safety interruption probability prediction method according to claim 1, characterized in that: include: A creation module, used to create a communication data set based on the collected target data, wherein the communication data set includes a training set; A training module, used for training the training set based on depthwise separable convolution and dilated convolution, and then applying a Transformer model to obtain a prediction model; The prediction module is used to predict the probability of safety interruption based on the prediction model.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the security interruption probability prediction method as described in any one of claims 1 to 5 is implemented.

Images