Mobile equipment owner authentication method based on de-noising and optimizing time domain convolutional network

Abstract

The invention belongs to the technical field of network security and deep learning, and discloses a mobile device owner authentication method based on denoising and optimizing a time domain convolutional network, which comprises the following steps: obtaining a training sample set which comprises a plurality of training samples, and each training sample covers feature data used by a mobile device in an owner authentication process, the labels correspond to the feature data; de-noising the training sample set by using a joint de-noising method; the de-noised training sample set is used to optimize and update an optimized time domain convolutional network to obtain an owner authentication model, and the optimized time domain convolutional network comprises a self-attention layer; and obtaining feature data of a to-be-authenticated mobile equipment user as input of the optimized and updated owner authentication model, and obtaining an authentication result about whether the to-be-authenticated mobile equipment user is the owner or not based on output of the owner authentication model. According to the invention, implicit, accurate and real-time identification of the mobile device can be realized.

Description

Technical Field

[0001] The present invention belongs to the field of network security and deep learning technology, and specifically relates to a method for authenticating the owner of a mobile device based on a denoising and optimized time-domain convolutional network. Background Art

[0002] The rise and popularity of mobile Internet is due to the rapid development of mobile communication technology and the Internet. With the growth of mobile device usage, more and more private data and personal identity authentication information are collected and used for the operation of mobile devices. Given the hardware, software and application scenarios of mobile devices, it is crucial to develop a suitable and reliable authentication model to protect user information security and prevent unauthorized access to local private and financial information stored on mobile devices. Implicit identity authentication collects and calculates data based on the perception of mobile terminals to complete the assessment of user authenticity. This process is based on the user's biometric data or behavioral data for identification.

[0003] At present, the global research on implicit real-time owner authentication for mobile devices is gradually developing towards real-time intelligence and integration. The specific technologies involved in this research include multi-sensor fusion, machine learning, mobile computing, human-computer interaction and other fields. At present, some progress has been made in the technology of behavior perception, data collection and behavior data processing. However, the current research is still in its early stages, mainly focusing on academic exploration and prototype verification. In the key technical direction, there is still a lack of a systematic research system, and there is also a large gap between it and practical applications.

[0004] In view of the above reasons, a more ideal and practical mobile device owner authentication technology must be proposed to solve the above problems. Summary of the invention

[0005] The purpose of the present invention is to provide a mobile device owner authentication method based on denoising and optimized time domain convolutional network to achieve implicit, accurate and real-time identification of mobile devices.

[0006] To achieve the above object, the technical solution adopted by the present invention is:

[0007] A method for authenticating the owner of a mobile device based on a denoising and optimized time-domain convolutional network, comprising:

[0008] Obtaining a training sample set, wherein the training sample set includes a plurality of training samples, each of which covers feature data used by the mobile device in the process of device owner authentication, and a label corresponding to the feature data;

[0009] The training sample set is denoised by a joint denoising method, wherein the joint denoising method comprises generating a training sample set by using a set empirical mode decomposition. Signal, Signal, Signal and signal, and use Criteria Extraction Signal and The details of the signal are analyzed using the Fast-ICA algorithm. The details of the signal, Signal, Signal and The details of the signal are de-noised and unmixed to obtain Quantity, Quantity, Quantity and Quantity, yes The components are denoised by wavelet threshold to obtain a denoised training sample set;

[0010] Optimizing and updating the optimized time-domain convolutional network using the denoising training sample set to obtain an owner authentication model, wherein the optimized time-domain convolutional network includes a self-attention layer;

[0011] The feature data of the user of the mobile device to be authenticated is obtained as the input of the optimized and updated machine owner authentication model, and the authentication result of whether the user of the mobile device to be authenticated is the machine owner is obtained based on the output of the machine owner authentication model.

[0012] Several optional methods are also provided below, but they are not intended to be additional limitations on the above-mentioned overall solution, but are merely further supplements or preferences. Under the premise that there are no technical or logical contradictions, each optional method can be combined with the above-mentioned overall solution separately, and multiple optional methods can also be combined.

[0013] Preferably, the step of obtaining a training sample set includes:

[0014] When the mobile device is in the process of owner authentication, collect three-axis data of the acceleration sensor, three-axis data of the gyroscope, and three-axis data of the gravity sensor;

[0015] Each data point collected is labeled with the user ID, sensor type, axis type, and the timestamp when the data point was collected.

[0016] Preferably, the step of denoising the training sample set using a joint denoising method comprises:

[0017] The three-axis data of the acceleration sensor, the three-axis data of the gyroscope and the three-axis data of the gravity sensor in the training sample set are taken to obtain a total of nine-axis time series data, and the joint denoising method is used to denoise the time series data of each axis.

[0018] Preferably, the structure of the optimized time-domain convolutional network includes an input layer, a time-domain convolutional layer, a self-attention layer and a fully connected layer connected in sequence.

[0019] Preferably, the structure of the time domain convolution layer includes multiple residual blocks, and a convolution layer and a ReLU activation function are added after every two residual blocks.

[0020] Preferably, the residual block includes a first branch and a second branch, the first branch includes a convolution layer, a ReLU activation function and a convolution layer in sequence, the second branch includes a downsampling layer, and the outputs of the first branch and the second branch are added as the output of the residual block.

[0021] Preferably, the attention value calculation process of the self-attention layer is as follows:

[0022] Calculate the correlation between the query vector and the key vector:

[0023] ;

[0024] In the formula, represents the query vector, represents the key vector element values, The query vector and the key vector are represented by The correlation of element values;

[0025] Compute the weights of a vector of values:

[0026] ;

[0027] In the formula, Indicates the value vector The weight of the elements, represents the SoftMax function, is the dimension of the key vector, The query vector and the key vector are represented by The correlation of element values;

[0028] Calculate the attention value:

[0029] ;

[0030] In the formula, is the attention value, Represents the value vector element value.

[0031] Preferably, the owner authentication model runs on a mobile device to perform local owner authentication, or the owner authentication model is deployed on a server to perform cloud-based owner authentication based on feature data uploaded by the mobile device.

[0032] Preferably, the output of the owner authentication model is the similarity between the user of the mobile device to be authenticated and the owner. If the similarity is greater than a threshold, the authentication result is consistent with the identity of the owner and the authentication is successful; otherwise, the authentication result is inconsistent with the identity of the owner and the authentication fails.

[0033] The present invention provides a mobile device owner authentication method based on denoising and optimized time-domain convolutional network. In order to realize implicit real-time owner authentication of mobile devices and put it into practical application, the present invention provides a mobile device owner authentication method based on denoising and optimized time-domain convolutional network. The method uses a joint denoising method to reduce the noise of the training sample set and introduces a self-attention mechanism into the time-domain convolutional network, thereby realizing implicit, accurate and real-time recognition of mobile devices.

[0034] The beneficial effects of the present invention are mainly reflected in: (1) By using the self-designed ICA-EEMD-wavelet threshold sensor signal joint denoising method to filter sensor noise, the influence of the training sample set on the results during the training phase is reduced, thereby improving the robustness of the model. (2) By using an optimized time domain convolutional network for efficient calculation, the relative relationship and local dependency between time series data are effectively captured by stacking convolutional layers and increasing the receptive field of the convolution kernel and the self-attention mechanism, deeper features are understood, the analysis ability of the model is improved, and the accuracy and feasibility of implicit owner authentication are enhanced. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 A flowchart of a method for authenticating the owner of a mobile device based on denoising and optimized time-domain convolutional networks according to the present invention;

[0036] Figure 2 is a flow chart of the combined denoising method of the present invention;

[0037] Figure 3 Schematic diagram of the structure of the time domain convolutional network optimized for the present invention. DETAILED DESCRIPTION

[0038] 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.

[0039] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0040] like Figure 1 As shown, this embodiment provides a method for authenticating the owner of a mobile device based on denoising and optimizing a time-domain convolutional network, comprising the following steps:

[0041] (1) Obtain a training sample set, where the training sample set includes multiple training samples, each of which covers feature data used by the mobile device in the owner authentication process and a label corresponding to the feature data.

[0042] When the mobile device is in the process of owner authentication, the three-axis data of the acceleration sensor, the three-axis data of the gyroscope and the three-axis data of the gravity sensor are collected, and a label is added to each collected data point, the label includes the user number, sensor type, axis type and timestamp when the data point is collected. This embodiment provides a specific collection method as follows:

[0043] (1-1) Preliminary preparation and parameter construction: Each sensor reading includes values ​​corresponding to the x, y, and z axes:

[0044] ;

[0045] ;

[0046] ;

[0047] In the formula, Indicates the first time steps, Indicates the acceleration sensor The data collected on the x-axis in time steps, Indicates the acceleration sensor The data collected on the y-axis at each time step, Indicates the acceleration sensor The data collected on the z-axis in time steps, Indicates the gyroscope The data collected on the x-axis in time steps, Indicates the gyroscope The data collected on the y-axis at each time step, Indicates the gyroscope The data collected on the z-axis in time steps, Indicates the gravity sensor The data collected on the x-axis in time steps, Indicates the gravity sensor The data collected on the y-axis at each time step, Indicates the gravity sensor The data collected on the z-axis is a time step.

[0048] Taking Android as an example, Android allows developers to refresh sensor data at fixed and customized intervals / delays after registering the sensor as a register listener(). There are four delays:

[0049] SENSOR_DELAY_FASTEST=0s;

[0050] SENSOR_DELAY_GAME=0:02s(50HZ);

[0051] SENSOR_DELAY_NORMAL=0.06s;

[0052] SENSOR_DELAY_UI=0.2s.

[0053] Considering battery consumption, data collection selects SENSOR_DELAY_GAME delay. This embodiment uses an existing mobile application to collect data, which performs continuous detection during the loading phase when the user actively uses the mobile device, and presents the characteristics of the user using the mobile device through sensor readings within a specified time.

[0054] Typically, the duration of the loading phase lasts 2 to 4 seconds. The duration is selected based on sensitivity analysis. When the time is less than 3 seconds, the final accuracy increases with the increase of time. When the time is greater than 3 seconds, the final accuracy decreases with the increase of time. This embodiment uses 3 seconds as the data collection duration.

[0055] (1-2) Data collection process: Taking Android as an example, the software uses the BroadcastReceiver() method to capture the system event of turning on the device's screen, and then starts a service that periodically queries the current application in the foreground. If the currently active application is different from the application in the last query, the software will recognize that a new application has started. If both conditions are met, data collection will last for 3 seconds:

[0056] The first condition is: the screen of the mobile device is turned on;

[0057] The second condition is: run a new application in the main page.

[0058] The boundaries of the 3D gravity sensor readings are as follows, and data that does not fit within this range will be discarded: .

[0059] (2) The training sample set is denoised using a joint denoising method. The joint denoising method includes generating Signal, Signal, Signal and signal, and use Criteria Extraction Signal and The details of the signal are analyzed using the Fast-ICA algorithm. The details of the signal, Signal, Signal and The details of the signal are de-noised and unmixed to obtain Quantity, Quantity, Quantity and Quantity, yes The components are denoised by wavelet threshold to obtain the denoised training sample set.

[0060] like Figure 2 As shown, the joint denoising method of ICA (Fast-ICA algorithm)-EEMD (Ensemble Empirical Mode Decomposition)-wavelet threshold proposed in this embodiment performs denoising on the data of each dimension, that is, the three-axis data of the acceleration sensor, the three-axis data of the gyroscope and the three-axis data of the gravity sensor in the training sample set are taken to obtain a total of nine-axis time series data, and the time series data of each axis is denoised by the joint denoising method. The specific denoising steps are as follows:

[0061] (2-1) The training sample set is decomposed into four layers of IMF signals using EEMD, and the source signal wave is found to be concentrated in Signal and In terms of signals, these two layers of IMF signals are retained. The EEMD decomposition process is as follows:

[0062] (2-1-1) Add equal-length Gaussian white noise to the original signal, and the new signal is as follows:

[0063] ;

[0064] In the formula, For the original signal Middle The signal after adding noise is For the The equal-length Gaussian white noise added is required to meet the following conditions:

[0065] ;

[0066] In the formula, is the standard deviation of the input signal, and are the intensity and number of times of adding noise respectively.

[0067] (2-1-2) Determine the mean by fitting the upper and lower envelopes of the signal When fitting, it is necessary to start from the preprocessed signal (the first time from step (2-1-1) is the signal In the subsequent steps, all local maximum points and local minimum points are identified in the detection signal obtained in step (2-1-3). Using the cubic spline curve interpolation method, all local maximum points and all local minimum points are connected respectively to obtain the upper envelope of the signal. and lower envelope The upper envelope is located above all local maximum points, while the lower envelope is located below all local minimum points. The calculation uses the following formula:

[0068] ;

[0069] In the formula, is the mean sequence.

[0070] (2-1-3) Remove the mean sequence from the preprocessed signal A detection signal is obtained, and it is determined whether the detection signal meets the IMF condition. If not, step (2-1-2) is repeated based on the detection signal until the detection signal meets the IMF condition, and the detection signal is output as the intrinsic mode function IMF component.

[0071] (2-1-4) Using signals The remaining symbols are calculated from the intrinsic mode functions as follows:

[0072] ;

[0073] In the formula, For the The residual signal after adding Gaussian white noise of equal length, For the The first time to add equal length Gaussian white noise The intrinsic mode function is For the The first intrinsic mode function after adding Gaussian white noise.

[0074] (2-1-5) Repeat steps (2-1-2) and (2-1-3) for the remaining signals to obtain the IMF components in sequence;

[0075] (2-1-6) Add white noise to the original signal again and repeat the above steps;

[0076] (2-1-7) The decomposition result of EEMD is obtained by lumping and averaging the IMF components of the same order obtained by decomposition:

[0077] ;

[0078] In the formula, For the The first time Gaussian white noise is added The intrinsic mode functions, is the total number of times Gaussian white noise is added.

[0079] (2-2) Utilization Guidelines for Signal and Extract the details of the signal. The criterion is based on repeated measurements of equal precision in normal distribution and is used when the interference or noise of singular data makes it difficult to satisfy the normal distribution. The specific steps are as follows:

[0080] (2-2-1) According to the principle of ensemble empirical mode decomposition, Signal and The signal satisfies the additive noise model, which is expressed as follows:

[0081] ;

[0082] ;

[0083] In the formula, express Signal, express Signal, express The remaining information after removing the noise from the signal. express The remaining information after removing the noise from the signal. express Noise in the signal, express The noise in the signal, and , .

[0084] (2-2-2) Based on Criteria, Noise Distribution satisfies ,in Indicates At each time point, the noise Falling in the range The probability of is 0.9973. Therefore, if The signal Signal at a time point The value does not fall within In, it is believed There must be significant errors in the Need to keep. Guidelines for The signal for detail extraction is expressed as:

[0085] ;

[0086] In the formula, for The details extracted from the signal.

[0087] (2-2-3) Noise variance of the signal for:

[0088] ;

[0089] In the formula, A function that returns the median, for High-frequency subband wavelet coefficients of the signal.

[0090] (2-2-4) Noise Distribution and Instead, using The criteria are:

[0091] ;

[0092] In the formula, for The details extracted from the signal.

[0093] (2-2-5) Noise variance of the signal for:

[0094] ;

[0095] In the formula, for High-frequency subband wavelet coefficients of the signal.

[0096] (2-3) The Fast-ICA algorithm is used to simultaneously perform noise reduction and demixing on the four IMF signals to obtain the best estimate signal of the source signal. The basic model of ICA is as follows:

[0097] (2-3-1) Assumption is a group dimensional independent source signals, yes dimensional observed signal, The components in Combined with each other, the matrix expression is as follows:

[0098] ;

[0099] In the formula, For unknown Mixing matrix, to meet the number of observation points Greater than or equal to the number of source signal points .

[0100] (2-3-2) The purpose of ICA analysis is to estimate the coefficient matrix without prior knowledge according to certain optimization criteria. The inverse matrix , by solving The inverse matrix of the independent source signal is obtained , that is to say, . The inverse matrix We can use the separation matrix Represented as follows:

[0101] ;

[0102] In the formula, is the source signal obtained by separation An approximate estimate of .

[0103] (2-4) Use wavelet soft threshold to further reduce the noise of IC1 reconstructed signal, and improve the noise reduction effect and performance indicators. The principle of wavelet threshold denoising is as follows: The basic principle of wavelet threshold denoising is to set a critical threshold , if the wavelet coefficient is less than , then the coefficient is mainly generated by noise, remove this part of the coefficient; if the wavelet coefficient is greater than or equal to , then the coefficient is mainly generated by the signal, retain this part of the coefficient, and finally perform inverse transformation on the processed wavelet coefficient to obtain the denoised signal. The steps of the wavelet threshold denoising method are as follows:

[0104] (2-4-1) Perform wavelet transform on the original signal of IC1 to the wavelet domain to obtain a set of wavelet decomposition coefficients.

[0105] (2-4-2) After threshold processing in the wavelet domain, a smaller wavelet coefficient is obtained to suppress the random noise. This embodiment uses a soft threshold method for denoising, and its threshold function is:

[0106] ;

[0107] In the formula, is the wavelet coefficient after threshold processing, is the symbolic function, is the threshold value, is the wavelet coefficient.

[0108] (2-4-3) The processed wavelet coefficients are used to reconstruct the signal to obtain the denoised signal. After denoising the 9-axis time series data, a denoised training sample set is obtained.

[0109] (3) The optimized time domain convolutional network is optimized and updated using the denoising training sample set to obtain the owner authentication model. The optimized time domain convolutional network includes a self-attention layer.

[0110] The time-domain convolutional network is a convolutional neural network-based architecture that effectively captures long-term dependencies in time series by stacking convolutional layers; the self-attention mechanism dynamically adjusts its weights according to the similarity between each time step in the input sequence. Figure 3 As shown, the implementation of the optimized time domain convolutional network in this embodiment is as follows: it includes an input layer, a time domain convolution layer, a self-attention layer and a fully connected layer connected in sequence. Among them, the structure of the time domain convolution layer includes multiple residual blocks (8 residual blocks are taken in this embodiment), and a convolution layer and a ReLU activation function are added after every two residual blocks. The residual block includes a first branch and a second branch, the first branch includes a convolution layer, a ReLU activation function and a convolution layer in sequence, and the second branch includes a downsampling layer (the downsampling layer can be replaced by a convolution layer as needed), and the outputs of the first branch and the second branch are added as the output of the residual block. The detailed description of each layer is as follows:

[0111] 1. Input layer: The input of the optimized time domain convolutional network is a tensor of shape (batch_size, input_channels, sequence_length), where batch_size is the batch size, input_channels is the number of input channels (or features), and sequence_length is the sequence length.

[0112] 2. Temporal convolution (TCN) layer: First, it passes through a series of residual blocks (ResidualBlock), each of which contains two one-dimensional convolutional layers (Conv1d) and a ReLU activation function. These convolutional layers may have different strides and dilation rates to capture features at different time scales. By introducing residual blocks, the network can learn the identity mapping of the input data, which helps alleviate the gradient vanishing problem in deep networks, allowing the network to be deeper and thus extract richer features.

[0113] 3. Convolutional layer: Each convolutional layer is followed by a ReLU activation function to increase nonlinearity so that the network can fit more complex functional relationships.

[0114] 4. Downsampling layer: If the stride is not equal to 1 or the dilation rate is greater than 1, the input is downsampled through a one-dimensional convolutional layer (downsample) to ensure dimensional matching during residual connection, which can ensure dimensional matching during residual connection, reduce the amount of calculation and improve model efficiency.

[0115] 5. Additional convolutional layers: After every two residual blocks, an additional one-dimensional convolutional layer with ReLU activation function is added, providing more flexibility to further extract features.

[0116] 6. Self-Attention Layer: After passing through a series of temporal convolutional layers, a self-attention layer is applied to capture the dependencies between different positions in the sequence. The self-attention mechanism can capture the dependencies between different positions in the sequence, which is particularly important for time series data. It allows the network to dynamically focus on important parts of the sequence, improving the expressiveness and generalization ability of the model.

[0117] Query vector (Query), key vector (Key), value vector (Value): First, the input is converted into query, key and value through three linear layers.

[0118] Attention score: Then, the dot product between the query vector and the key vector is calculated to get the attention score. That is, the correlation between the query vector and the key vector is calculated:

[0119] ;

[0120] In the formula, represents the query vector, represents the key vector element values, The first The correlation of element values.

[0121] Attention weight: Apply the Softmax function to the attention score to get the attention weight. Introduce a calculation method similar to SoftMax to convert the scores of the first stage into numerical values. On the one hand, the original calculated scores can be normalized to form a probability distribution where the sum of the weights of all elements is 1; on the other hand, the weights of important elements can be highlighted through the internal mechanism of SoftMax. That is, the following formula is generally used:

[0122] ;

[0123] In the formula, Indicates the value vector The weight of the elements, represents the SoftMax function, is the dimension of the key vector, The query vector and the key vector are represented by The correlation of element values.

[0124] Weighted Sum: Use the attention weights to perform weighted summation on the values ​​to obtain a weighted feature representation.

[0125] Aggregation: Finally, the weighted feature representation is summed over the sequence length dimension to aggregate information, which helps integrate the global information in the sequence into the representation of each time step. That is, the Value is weighted and summed according to the weight coefficient to obtain the final attention value:

[0126] ;

[0127] In the formula, is the attention value, Represents the value vector element value.

[0128] 7. Fully Connected Layer Finally, a fully connected layer is used to map the output of the self-attention layer to the probability distribution of the target category. The input size of this fully connected layer is input_channels, and the output size is num_classes, which is the number of categories. The output of the time domain convolutional network is a tensor of shape (batch_size, num_classes), which represents the probability that each sample belongs to each category.

[0129] (4) Obtaining feature data of the user of the mobile device to be authenticated as input to the optimized and updated machine owner authentication model, and obtaining an authentication result of whether the user of the mobile device to be authenticated is the machine owner based on the output of the machine owner authentication model.

[0130] The characteristic data of the user of the mobile device to be authenticated is obtained as the input of the owner authentication model. During the verification process, this embodiment uses the existing IOS development library AFNetworking and the existing ApacheHttpClient library of the JAVA application to cover mobile devices of different operating systems, so as to simplify the network request operation when interacting with the server. The data sent by the client is processed by the server built on Django. The C++ data preparation module is used to distinguish between two motion states and filter out invalid data for the fingerprint user mode. LibXtract is a function library for audio feature extraction. The library implements a variety of statistical feature extraction functions, such as variance, mean deviation, skew and angularization, etc. These functions can share calculation results to improve efficiency, reduce repeated calculations, improve processing speed, and increase the flexibility of the library, allowing developers to freely combine different feature extraction functions as needed. Considering the versatility of the extraction functions in the library, the system uses LibXtract to extract features. In the case of local identity authentication, identity authentication is performed by using Libsvm for iOS and AndroidLibsvm for Android. In the case of server-side identity authentication, the collected samples are uploaded to the cloud server for judgment, and the cloud server returns the judgment similarity. If the similarity is lower than the set threshold, it can be considered that the actual user does not match the identity of the owner. When the output of the owner authentication model is the similarity between the user of the mobile device to be authenticated and the owner, if the similarity is greater than the threshold, the authentication result is consistent with the identity of the owner and the authentication is successful; otherwise, the authentication result is inconsistent with the identity of the owner and the authentication fails.

[0131] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0132] The above-mentioned embodiments only express several implementation modes of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the attached claims.

Claims

1. A method for authenticating the owner of a mobile device based on denoising and optimized time-domain convolutional networks, characterized in that: The mobile device owner authentication method based on denoising and optimizing time-domain convolutional network includes: Obtaining a training sample set, wherein the training sample set includes a plurality of training samples, each of which covers feature data used by the mobile device in the process of device owner authentication, and a label corresponding to the feature data; The training sample set is denoised by a joint denoising method, wherein the joint denoising method comprises generating a training sample set by using a set empirical mode decomposition. Signal, Signal, Signal and signal, and use Criteria Extraction Signal and The details of the signal are analyzed using the Fast-ICA algorithm. The details of the signal, Signal, Signal and The details of the signal are de-noised and unmixed to obtain Quantity, Quantity, Quantity and Quantity, yes The components are denoised by wavelet threshold to obtain a denoised training sample set; Optimizing and updating the optimized time-domain convolutional network using the denoising training sample set to obtain an owner authentication model, wherein the optimized time-domain convolutional network includes a self-attention layer; The feature data of the user of the mobile device to be authenticated is obtained as the input of the optimized and updated machine owner authentication model, and the authentication result of whether the user of the mobile device to be authenticated is the machine owner is obtained based on the output of the machine owner authentication model.

2. The method for mobile device owner authentication based on denoising and optimized time-domain convolutional network according to claim 1, characterized in that: The step of obtaining a training sample set includes: When the mobile device is in the process of owner authentication, collect three-axis data of the acceleration sensor, three-axis data of the gyroscope, and three-axis data of the gravity sensor; Each data point collected is labeled with the user ID, sensor type, axis type, and the timestamp when the data point was collected.

3. The method for authenticating the owner of a mobile device based on denoising and optimized time-domain convolutional network according to claim 2, characterized in that: The method of denoising the training sample set by using a joint denoising method comprises: The three-axis data of the acceleration sensor, the three-axis data of the gyroscope and the three-axis data of the gravity sensor in the training sample set are taken to obtain a total of nine-axis time series data, and the joint denoising method is used to denoise the time series data of each axis.

4. The method for mobile device owner authentication based on denoising and optimized time-domain convolutional network according to claim 1, characterized in that: The structure of the optimized time-domain convolutional network includes an input layer, a time-domain convolutional layer, a self-attention layer and a fully connected layer connected in sequence.

5. The method for authenticating the owner of a mobile device based on denoising and optimized time-domain convolutional network according to claim 4, characterized in that: The structure of the time domain convolution layer includes multiple residual blocks, and a convolution layer and a ReLU activation function are added after every two residual blocks.

6. The method for mobile device owner authentication based on denoising and optimized time-domain convolutional network according to claim 5, characterized in that: The residual block includes a first branch and a second branch, the first branch includes a convolution layer, a ReLU activation function and a convolution layer in sequence, the second branch includes a downsampling layer, and the outputs of the first branch and the second branch are added as the output of the residual block.

7. The method for mobile device owner authentication based on denoising and optimized time-domain convolutional network according to claim 4, characterized in that: The attention value calculation process of the self-attention layer is as follows: Calculate the correlation between the query vector and the key vector: ； In the formula, represents the query vector, represents the key vector element values, The query vector and the key vector are represented by The correlation of element values; Compute the weights of a vector of values: ； In the formula, Indicates the value vector The weight of the elements, represents the SoftMax function, is the dimension of the key vector, The query vector and the key vector are represented by The correlation of element values; Calculate the attention value: ； In the formula, is the attention value, Represents the value vector element value.

8. The method for mobile device owner authentication based on denoising and optimized time-domain convolutional network according to claim 1, characterized in that: The machine owner authentication model runs on a mobile device to perform local machine owner authentication, or the machine owner authentication model is deployed on a server side to perform cloud-based machine owner authentication based on feature data uploaded by the mobile device.

9. The method for mobile device owner authentication based on denoising and optimized time-domain convolutional network according to claim 1, characterized in that: The output of the owner authentication model is the similarity between the user of the mobile device to be authenticated and the owner. If the similarity is greater than a threshold, the authentication result is consistent with the identity of the owner and the authentication is successful; otherwise, the authentication result is inconsistent with the identity of the owner and the authentication fails.

Images