CN118468327B

CN118468327B - Sensitive information feature extraction method, device, electronic device and medium

Info

Publication number: CN118468327B
Application number: CN202310520275.7A
Authority: CN
Inventors: 苏义军; 李叶昕; 王昌钰; 张钧波; 郑宇�; 陈高德
Original assignee: Jingdong City Beijing Digital Technology Co Ltd
Current assignee: Jingdong City Beijing Digital Technology Co Ltd
Priority date: 2023-05-09
Filing date: 2023-05-09
Publication date: 2025-02-21
Anticipated expiration: 2043-05-09
Also published as: CN118468327A

Abstract

The embodiment of the invention discloses a sensitive information feature extraction method, a sensitive information feature extraction device, electronic equipment and a medium. The method comprises the steps of carrying out information preprocessing on sensitive information of each user to generate preprocessed user information to obtain preprocessed user information sets, carrying out information preprocessing on scene description information of each scene to generate preprocessed scene information to obtain preprocessed scene information sets, extracting characteristic information of each preprocessed user information to generate user characteristic information to obtain user characteristic information sets, extracting characteristic information of each preprocessed scene information to obtain scene characteristic information sets, and generating information characteristic sets according to the user characteristic information sets and the scene characteristic information sets. This embodiment is related to artificial intelligence, and feature extraction for user sensitive information and scene description information can be achieved more accurately.

Description

Sensitive information feature extraction method and device, electronic equipment and medium

Technical Field

Embodiments of the present disclosure relate to the field of computer technologies, and in particular, to a method, an apparatus, an electronic device, and a medium for extracting features of sensitive information.

Background

At present, in the big data age, the leakage of user sensitive information can bring serious security accidents. Protection of sensitive information of the user's individual is an important concern for current people. The identification and classification of sensitive information are the precondition of personal information security protection. Aiming at downstream tasks such as identification, classification, grading and the like of sensitive information, most of the prior art respectively carries out different model designs on different downstream tasks. However, these downstream tasks based on sensitive information all need to process multi-modal sensitive information, and for feature extraction of user sensitive information, a method is generally adopted in which the user sensitive information is directly input into a feature extraction model to generate user feature information.

However, the inventors have found that when the above manner is adopted to extract features, there are often the following technical problems:

aiming at the multi-mode form of the user sensitive information, the feature extraction model can only extract the single-mode user feature information, and the extracted user feature information is not comprehensive enough, so that the extraction of the user feature information is not accurate and efficient enough.

The above information disclosed in this background section is only for enhancement of understanding of the background of the inventive concept and, therefore, may contain information that does not form the prior art that is already known to those of ordinary skill in the art in this country.

Disclosure of Invention

The disclosure is in part intended to introduce concepts in a simplified form that are further described below in the detailed description. The disclosure is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter.

Some embodiments of the present disclosure propose sensitive information feature extraction methods, apparatuses, electronic devices, computer readable media and program products to solve the technical problems mentioned in the background section above.

In a first aspect, some embodiments of the present disclosure provide a method for extracting a feature of sensitive information, including performing information preprocessing on each user sensitive information in a sample data set to generate preprocessed user information, obtaining a preprocessed user information set, performing information preprocessing on each scene description information in a corresponding scene description information set to generate preprocessed scene information, obtaining a preprocessed scene information set, wherein the sample data set includes user sensitive information in a multi-modal form, extracting feature information of each preprocessed user information in the preprocessed user information set to generate user feature information to obtain a user feature information set, extracting feature information of each preprocessed scene information in the preprocessed scene information set according to the user feature information set to generate scene feature information to obtain a scene feature information set, and generating an information feature set according to the user feature information set and the scene feature information set.

Optionally, the user sensitive information in the sample data set includes user attribute information and user attribute content information, and the performing information preprocessing on each user sensitive information in the sample data set to generate preprocessed user information includes performing information preprocessing on the user attribute information included in the user sensitive information to obtain preprocessed user attribute information, performing information preprocessing on the user attribute content information included in the user sensitive information to obtain preprocessed attribute content information, and generating preprocessed user information for the user sensitive information according to the preprocessed user attribute information and the preprocessed attribute content information.

Optionally, extracting the feature information of each piece of the preprocessed user information in the preprocessed user information set to generate the user feature information includes extracting attribute content feature information corresponding to preprocessed attribute content information included in the preprocessed user information as the user feature information.

Optionally, the method comprises the steps of extracting feature information of each piece of preprocessing scene information in the preprocessing scene information set according to the user feature information set to generate scene feature information, wherein the feature information comprises the steps of determining user feature information corresponding to the preprocessing scene information as target user feature information, inputting preprocessing user attribute information corresponding to the target user feature information and the preprocessing scene information into a first scene feature information generation model based on an attention mechanism to generate first initial scene feature information, and generating scene feature information aiming at the preprocessing scene information according to the first initial scene feature information.

The method comprises the steps of generating scene characteristic information for the preprocessing scene information according to the first initial scene characteristic information, wherein the scene characteristic information comprises the steps of carrying out information combination on the target user characteristic information and the preprocessing scene information to obtain first combination information, inputting the first combination information into a second scene characteristic information generation model to generate second initial scene characteristic information, and carrying out information combination on the first initial scene characteristic information, the second initial scene characteristic information and the preprocessing scene information to obtain second combination information serving as the scene characteristic information for the preprocessing scene information.

Optionally, the extracting attribute content feature information corresponding to the preprocessing attribute content information included in the preprocessing user information, as the user feature information, includes, in response to determining that a modality form of user sensitive information corresponding to the preprocessing user information is a text form, inputting the preprocessing user information into a bidirectional time sequence neural network model to generate hidden layer output information for each moment, and performing weighted summation processing on each obtained hidden layer output information by using a first attention mechanism model to generate attribute content feature information as the user feature information.

Optionally, the extracting attribute content feature information corresponding to the preprocessing attribute content information included in the preprocessing user information as the user feature information includes performing multidimensional convolution on the preprocessing user information in response to determining that a mode form of user sensitive information corresponding to the preprocessing user information is an image form, obtaining a plurality of convolution results, splicing each convolution result in the plurality of convolution results to obtain a splicing result, and inputting the splicing result into an image feature information generation model to generate attribute content feature information corresponding to the preprocessing attribute content information as the user feature information.

The method comprises the steps of extracting attribute content characteristic information corresponding to preprocessing attribute content information included in the preprocessing user information, responding to the fact that a mode form of user sensitive information corresponding to the preprocessing user information is determined to be a video form, carrying out multi-dimensional convolution on each frame image in a frame image sequence corresponding to the preprocessing user information to obtain a plurality of convolution results, splicing all convolution results in the plurality of convolution results to obtain a splicing result, inputting the splicing result into an image characteristic information generation model to generate the attribute content characteristic information corresponding to the preprocessing attribute content information to serve as target attribute content characteristic information, inputting the obtained target attribute content characteristic information sequence into a time sequence neural network model to output hidden layer output result sets at all moments, and carrying out weighted summation processing on all hidden layer output information in the hidden layer output result sets by using a first attention mechanism model to generate the attribute content characteristic information as the user characteristic information.

Optionally, the extracting the attribute content feature information corresponding to the preprocessing attribute content information included in the preprocessing user information as the user feature information includes generating attribute content feature information for the preprocessing user information as the user feature information by using an audio feature information generation model in response to determining that a mode form of user sensitive information corresponding to the preprocessing user information is an audio form.

Optionally, the method further comprises the step of inputting the information feature set into a classification model corresponding to the target classification task to generate a classification result.

In a second aspect, some embodiments of the present disclosure provide a sensitive information feature extraction apparatus, including a preprocessing unit configured to perform information preprocessing on each user sensitive information in a sample data set to generate preprocessed user information, to obtain preprocessed user information sets, and perform information preprocessing on each scene description information in a corresponding scene description information set to generate preprocessed scene information, to obtain preprocessed scene information sets, wherein the sample data set includes user sensitive information in a multi-modal form, a first extraction unit configured to extract feature information of each preprocessed user information in the preprocessed user information sets to generate user feature information, to obtain a user feature information set, a second extraction unit configured to extract feature information of each preprocessed scene information in the preprocessed scene information sets according to the user feature information sets, to generate scene feature information, to obtain scene feature information sets, and a generation unit configured to generate an information feature set according to the user feature information sets and the scene feature information sets.

Optionally, the user sensitive information in the sample data set comprises user attribute information and user attribute content information, and the preprocessing unit can be configured to perform information preprocessing on the user attribute information included in the user sensitive information to obtain preprocessed user attribute information, perform information preprocessing on the user attribute content information included in the user sensitive information to obtain preprocessed attribute content information, and generate preprocessed user information for the user sensitive information according to the preprocessed user attribute information and the preprocessed attribute content information.

Alternatively, the first extraction unit may be configured to extract, as the user feature information, attribute content feature information corresponding to the preprocessing attribute content information included in the preprocessing user information described above.

Alternatively, the second extraction unit may be configured to determine user feature information corresponding to the above-mentioned preprocessed scene information as target user feature information, input preprocessed user attribute information corresponding to the above-mentioned target user feature information and the above-mentioned preprocessed scene information to a first scene feature information generation model based on an attention mechanism to generate first initial scene feature information, and generate scene feature information for the above-mentioned preprocessed scene information based on the above-mentioned first initial scene feature information.

Alternatively, the second extraction unit may be configured to combine the target user feature information and the preprocessing scene information to obtain first combined information, input the first combined information to a second scene feature information generation model to generate second initial scene feature information, and combine the first initial scene feature information, the second initial scene feature information and the preprocessing scene information to obtain second combined information as scene feature information for the preprocessing scene information.

Optionally, the first extraction unit may be configured to input the preprocessed user information to the bidirectional time-sequential neural network model to generate hidden layer output information for each moment in response to determining that the modality form of the user sensitive information corresponding to the preprocessed user information is text form, and perform weighted summation processing on each obtained hidden layer output information by using the first attention mechanism model to generate attribute content feature information as the user feature information.

Optionally, the first extraction unit may be configured to perform multidimensional convolution on the preprocessed user information to obtain a plurality of convolution results in response to determining that the modality form of the user sensitive information corresponding to the preprocessed user information is an image form, splice each convolution result in the plurality of convolution results to obtain a splice result, and input the splice result into an image feature information generation model to generate attribute content feature information corresponding to the preprocessed attribute content information as the user feature information.

Optionally, the first extraction unit may be configured to perform, in response to determining that the modality of the user sensitive information corresponding to the preprocessed user information is in a video form, a processing step of performing multi-dimensional convolution on the image to obtain a plurality of convolution results for each frame image in the frame image sequence corresponding to the preprocessed user information, performing stitching on each convolution result in the plurality of convolution results to obtain a stitching result, inputting the stitching result to an image feature information generation model to generate attribute content feature information corresponding to the preprocessed attribute content information as target attribute content feature information, inputting the obtained target attribute content feature information sequence to a time-series neural network model to output a hidden layer output result set at each moment, and performing weighted summation processing on each hidden layer output information in the hidden layer output result set by using a first attention mechanism model to generate attribute content feature information as user feature information.

Alternatively, the first extraction unit may be configured to generate attribute content feature information for the pre-processed user information as the user feature information using an audio feature information generation model in response to determining that the modality form of the user sensitive information corresponding to the pre-processed user information is an audio form.

Optionally, the device further comprises inputting the information feature set into a classification model corresponding to the target classification task to generate a classification result.

In a third aspect, some embodiments of the present disclosure provide an electronic device comprising one or more processors, a storage device having one or more programs stored thereon, which when executed by the one or more processors, cause the one or more processors to implement a method as described in any of the implementations of the first aspect.

In a fourth aspect, some embodiments of the present disclosure provide a computer readable medium having a computer program stored thereon, wherein the program when executed by a processor implements a method as described in any of the implementations of the first aspect.

In a fifth aspect, some embodiments of the present disclosure provide a computer program product comprising a computer program which, when executed by a processor, implements the method described in any of the implementations of the first aspect above.

The above embodiments of the present disclosure have the advantage that feature extraction for user sensitive information and scene description information can be more accurately achieved by the sensitive information feature extraction method of some embodiments of the present disclosure. Specifically, the reason for the insufficient accuracy of the related classification results is that the feature extraction model can only extract single-mode user feature information aiming at the multi-mode form of the user sensitive information, and the extracted user feature information is insufficient, so that the extraction of the user feature information is not accurate and efficient. Based on this, in the sensitive information feature extraction method of some embodiments of the present disclosure, first, information preprocessing is performed on each user sensitive information in a sample data set to generate preprocessed user information, so as to obtain a preprocessed user information set, and information preprocessing is performed on each scene description information in a corresponding scene description information set, so as to generate preprocessed scene information, so as to obtain a preprocessed scene information set. Wherein the sample dataset comprises user-sensitive information in a multi-modal form. Here, the feature extraction of the subsequent information is facilitated by preprocessing the user-sensitive information and the scene description information. Then, the feature information of each piece of the preprocessed user information in the preprocessed user information set can be accurately extracted to generate user feature information, so as to obtain a user feature information set. Here, the extracted user feature information set facilitates the generation of a subsequent information feature set. Furthermore, according to the user characteristic information set, the characteristic information of each piece of preprocessing scene information in the preprocessing scene information set can be accurately extracted to generate scene characteristic information, so as to obtain the scene characteristic information set. Here, the resulting set of scene feature information is used for the generation of a subsequent set of training samples. And finally, generating an information feature set according to the user feature information set and the scene feature information set. Here, by extracting a feature information set (i.e., a scene feature information set) of a scene information set corresponding to the sample data set, an information feature set with more abundant feature information can be obtained.

Drawings

The above and other features, advantages, and aspects of embodiments of the present disclosure will become more apparent by reference to the following detailed description when taken in conjunction with the accompanying drawings. The same or similar reference numbers will be used throughout the drawings to refer to the same or like elements. It should be understood that the figures are schematic and that elements and components are not necessarily drawn to scale.

1-2 Are schematic diagrams of one application scenario of a sensitive information feature extraction method according to some embodiments of the present disclosure;

FIG. 3 is a flow chart of some embodiments of a sensitive information feature extraction method according to the present disclosure;

FIG. 4 is a flow chart of further embodiments of a sensitive information feature extraction method according to the present disclosure;

FIG. 5 is a schematic structural view of some embodiments of a sensitive information feature extraction device according to the present disclosure;

fig. 6 is a schematic structural diagram of an electronic device suitable for use in implementing some embodiments of the present disclosure.

Detailed Description

Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While certain embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete. It should be understood that the drawings and embodiments of the present disclosure are for illustration purposes only and are not intended to limit the scope of the present disclosure.

It should be noted that, for convenience of description, only the portions related to the present invention are shown in the drawings. Embodiments of the present disclosure and features of embodiments may be combined with each other without conflict.

It should be noted that the terms "first," "second," and the like in this disclosure are merely used to distinguish between different devices, modules, or units and are not used to define an order or interdependence of functions performed by the devices, modules, or units.

It should be noted that references to "one", "a plurality" and "a plurality" in this disclosure are intended to be illustrative rather than limiting, and those of ordinary skill in the art will appreciate that "one or more" is intended to be understood as "one or more" unless the context clearly indicates otherwise.

The names of messages or information interacted between the various devices in the embodiments of the present disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

Operations such as collection, storage, use, etc. of user information (e.g., user sensitive information, scene description information, personal privacy information) referred to in the present disclosure, before performing the corresponding operations, the relevant organization or individual is exhausted to the end including developing personal information security impact assessment, fulfilling obligations to the personal information body, soliciting authorized consent from the personal information body in advance, etc.

The present disclosure will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.

Fig. 1-2 are schematic diagrams of an application scenario of a sensitive information feature extraction method according to some embodiments of the present disclosure.

In the application scenario of fig. 1-2, first, the electronic device 101 may perform information preprocessing on each user sensitive information in the sample data set 102 to generate preprocessed user information, to obtain a preprocessed user information set 104, and perform information preprocessing on each scene description information in the corresponding scene description information set 103 to generate preprocessed scene information, to obtain a preprocessed scene information set 105. Wherein the sample dataset 102 comprises user-sensitive information in multimodal form. In the present application scenario, sample data set 102 may include first user-sensitive information 1021, second user-sensitive information 1022, and third user-sensitive information 1023. Specifically, the first user sensitive information 1021 corresponds to a modality form that is an image form. The second user sensitive information 1022 corresponds to the modality being in audio form. Third user-sensitive information 1023 corresponds to a modality in the form of video. The set of preprocessed user information 104 may include first preprocessed user information 1041 corresponding to the first user sensitive information 1021, second preprocessed user information 1042 corresponding to the second user sensitive information 1022, and third preprocessed user information 1043 corresponding to the third user sensitive information 1023. The scene description information set 103 may include first scene description information 1031, second scene description information 1032, and third scene description information 1033. The set of pre-processed scene information 105 may include first pre-processed scene information 1051 corresponding to the first scene description information 1031, second pre-processed scene information 1052 corresponding to the second scene description information 1032, and third pre-processed scene information 1053 corresponding to the third scene description information 1033. The electronic device 101 may then extract the feature information of each of the preprocessed user information sets 104 described above to generate user feature information, resulting in a user feature information set 106. In the present application scenario, the user characteristic information set 106 may include first user characteristic information 1061 corresponding to the first user sensitive information 1021, second user characteristic information 1062 corresponding to the second user sensitive information 1022, and third user characteristic information 1063 corresponding to the third user sensitive information 1023. Next, the electronic device 101 may extract, according to the user feature information set 106, feature information of each of the above-described preprocessed scene information sets 105 to generate scene feature information, resulting in a scene feature information set 107. In the present application scenario, the scene feature information set 107 may include first scene feature information 1071 corresponding to the first scene description information 1031, second scene feature information 1072 corresponding to the second scene description information 1032, and third scene feature information 1073 corresponding to the third scene description information 1033. Finally, the electronic device 101 may generate the information feature set 108 from the user feature information set 106 and the scene feature information set 107. in the present application scenario, the information feature set 108 includes a first information feature 1081 for the first scene feature information 1071 and the first user feature information 1061, a second information feature 1082 for the second scene feature information 1072 and the second user feature information 1062, and a third information feature 1083 for the third scene feature information 1073 and the third user feature information 1063.

Note that, the electronic device 101 may be hardware, or may be software. When the electronic device is hardware, the electronic device may be implemented as a distributed cluster formed by a plurality of servers or terminal devices, or may be implemented as a single server or a single terminal device. When the electronic device is embodied as software, it may be installed in the above-listed hardware device. It may be implemented as a plurality of software or software modules, for example, for providing distributed services, or as a single software or software module. The present invention is not particularly limited herein.

It should be understood that the number of electronic devices in fig. 1-2 is merely illustrative. There may be any number of electronic devices as desired for an implementation.

With continued reference to fig. 3, a flow 300 of some embodiments of a sensitive information feature extraction method according to the present disclosure is shown. The sensitive information characteristic extraction method comprises the following steps:

Step 301, performing information preprocessing on each user sensitive information in the sample data set to generate preprocessed user information, thereby obtaining a preprocessed user information set, and performing information preprocessing on each scene description information in the corresponding scene description information set to generate preprocessed scene information, thereby obtaining a preprocessed scene information set.

In some embodiments, the execution body of the above-mentioned sensitive information feature extraction method (for example, the electronic device 101 shown in fig. 1) may perform information preprocessing on each user sensitive information in the sample data set to generate preprocessed user information, to obtain a preprocessed user information set, and perform information preprocessing on each scene description information in the corresponding scene description information set to generate preprocessed scene information, to obtain a preprocessed scene information set. Wherein the sample dataset comprises user-sensitive information in a multi-modal form. I.e. the individual user sensitive information in the sample data set may be user sensitive information of various modalities. For example, the various modalities may be image modalities, audio modalities, text modalities, and video modalities. The user sensitive information may be various aspects of the user. In practice, the user sensitive information may include personal privacy information of the user. For example, the personal privacy information may be user taxi taking data. The scene description information in the scene description information set and the user sensitive information in the sample data set have a one-to-one correspondence. The scene description information may be scene information describing a scene where the user-sensitive information corresponds to an occurrence scene. Wherein the scene description information may be information in the form of text.

For example, for privacy data in which the user sensitive information is in a text mode, the executing body may use a word segmentation tool to segment words of the user sensitive information to obtain a word set. Then, the words in the Word set are encoded through Word embedding tool Word2Vec to obtain encoding vectors as preprocessing user information. Aiming at privacy data of which the user sensitive information is in an image mode, the execution main body can perform standardized processing on the user sensitive information to obtain a standardized result. And then, readjusting the size corresponding to the standardized result, and cutting to obtain a cutting result which is used as the preprocessing user information.

As yet another example, the execution subject may segment the scene description information to obtain a scene word set. Then, the scene words in the scene Word set are encoded through a Word embedding tool Word2Vec, and a scene encoding vector set is obtained. And finally, combining the scene coding vector sets to obtain combined information serving as preprocessing scene information.

In some alternative implementations of some embodiments, the user-sensitive information in the sample dataset includes user attribute information and user attribute content information. Wherein the user attribute information may be respective attribute names involved in the user sensitive information. The user attribute content information may be attribute content corresponding to each attribute name in the user sensitive information.

Optionally, the information preprocessing for each user sensitive information in the sample data set to generate preprocessed user information may include the following steps:

The first step, the executing body may perform information preprocessing on user attribute information included in the user sensitive information to obtain preprocessed user attribute information, and perform information preprocessing on user attribute content information included in the user sensitive information to obtain preprocessed attribute content information.

Specifically, for privacy data in which the user attribute information is in a text mode, the execution body may use a word segmentation tool to segment the user attribute information to obtain a word set. Then, the words in the Word set are encoded through Word embedding tool Word2Vec to obtain encoding vectors which are used as preprocessing user attribute information. Similarly, for privacy data in which the user attribute content information is in a text mode, the user attribute content information can be subjected to information preprocessing in the same way, and detailed description is omitted.

And a second step, the executing body can generate the preprocessing user information aiming at the user sensitive information according to the preprocessing user attribute information and the preprocessing attribute content information.

As an example, the execution subject may perform information fusion of the preprocessing user attribute information and the preprocessing attribute content information to generate fusion information as preprocessing user information for the user sensitive information.

Step 302, extracting the feature information of each piece of pre-processing user information in the pre-processing user information set to generate user feature information, thereby obtaining a user feature information set.

In some embodiments, the executing entity may extract feature information of each piece of preprocessed user information in the preprocessed user information set to generate user feature information, and obtain the user feature information set. The user characteristic information can represent characteristic information corresponding to the user sensitive information, namely, characteristic content corresponding to the user sensitive information is reflected.

As an example, for privacy data in which the user sensitive information is in a text mode, the executing entity may input corresponding preprocessed user information into the Bert (Bidirectional Encoder Representations from Transformers) coding model to output user feature information. For privacy data in which the user sensitive information is in an image mode, the execution subject can input corresponding preprocessed user information into a multi-layer serial connected convolutional neural network model to generate user characteristic information.

In some optional implementations of some embodiments, the executing body may extract attribute content feature information corresponding to the preprocessing attribute content information included in the preprocessing user information as the user feature information.

Optionally, the extracting attribute content feature information corresponding to the preprocessing attribute content information included in the preprocessing user information may include the following steps as the user feature information:

In response to determining that the mode form of the user sensitive information corresponding to the preprocessed user information is a video form, the following processing steps are executed for each frame image in the frame image sequence corresponding to the preprocessed user information:

and 1, performing multidimensional convolution on the image to obtain a plurality of convolution results.

As an example, the execution body may perform multi-dimensional convolution on the image using a plurality of convolution kernels of different dimensions to generate a plurality of convolution results.

And 2, splicing all the convolution results in the plurality of convolution results to obtain a spliced result.

And 3, inputting the splicing result into an image characteristic information generation model to generate attribute content characteristic information corresponding to the preprocessing attribute content information as target attribute content characteristic information. Wherein the image characteristic information generation model may be a neural network model that generates image characteristic information (i.e., image characteristic information in the form of a vector). In practice, the image characteristic information generation model may be a multi-layer serial-connected full-connection (Fully Connected layers, FC) model.

And secondly, inputting the obtained target attribute content characteristic information sequence into a time sequence neural network model to output hidden layer output result sets at all times. In practice, the time series neural network model may be a Long Short Term Memory network (LSTM) model. Each time in each time has a one-to-one correspondence with the hidden layer output result in the hidden layer output result set. The hidden layer output result is not described in detail.

And thirdly, carrying out weighted summation processing on each hidden layer output information in the hidden layer output result set by using a first attention mechanism model so as to generate attribute content characteristic information serving as user characteristic information. Wherein the first attention mechanism model may be based on a neural network model of the attention mechanism. In practice, the first attention mechanism model may be a multi-headed attention mechanism model.

As an example, first, the above-described execution subject may generate a weight for each hidden layer output information using the first attention mechanism model. And multiplying Mi Geyin layers of output information in each hidden layer of output information with the corresponding weight to obtain each multiplication result. And finally, adding the multiplication results to obtain an addition result which is used as the user characteristic information.

In some optional implementations of some embodiments, in response to determining that the modality form of the user sensitive information corresponding to the preprocessed user information is an audio form, the execution body may generate attribute content feature information for the preprocessed user information as the user feature information using an audio feature information generation model. Wherein the audio feature information generation model may be a model that generates audio feature information. In practice, the modality form for the corresponding user sensitive information is an audio form, and the pre-processed user information may be a mel-graph. The audio feature information generation model may be a multi-layer Residual Network (ResNet).

Step 303, extracting feature information of each piece of preprocessing scene information in the preprocessing scene information set according to the user feature information set to generate scene feature information, and obtaining a scene feature information set.

In some embodiments, the executing body may extract feature information of each piece of preprocessing scene information in the preprocessing scene information set according to the user feature information set, so as to generate scene feature information, and obtain the scene feature information set. The scene characteristic information may represent characteristic information of scene content corresponding to the scene description information.

For example, for each piece of preprocessing scene information, first, the execution subject may splice the preprocessing scene information and the corresponding user feature information to obtain splice information. The stitching information is then input to a recurrent neural network (Recurrent Neural Network, RNN) to generate scene characteristic information.

In some optional implementations of some embodiments, extracting feature information of each piece of preprocessed scene information in the preprocessed scene information set according to the user feature information set to generate scene feature information may include the steps of:

In the first step, the execution subject may determine user characteristic information corresponding to the preprocessing scene information as target user characteristic information.

In the second step, the execution subject may input the preprocessed user attribute information and the preprocessed scene information corresponding to the target user feature information into a first scene feature information generation model based on an attention mechanism to generate first initial scene feature information. Wherein the first scene feature information generation model may be a model that generates scene feature information. In practice, the first scene feature information generation model based on the attention mechanism may be a multi-headed attention mechanism neural network model.

As an example, the execution subject may use the word segmentation set corresponding to the preprocessed user attribute information as a value (value) in the first scene feature information generation model, and use the word segmentation set corresponding to the preprocessed scene information as a key (key) and query information (query) in the first scene feature information generation model to generate the first initial scene feature information.

Third, generating scene characteristic information for the preprocessing scene information according to the first initial scene characteristic information.

As an example, the execution subject may determine the first initial scene feature information as scene feature information for the preprocessing scene information.

In some optional implementations of some embodiments, generating the scene feature information for the preprocessing scene information according to the first initial scene feature information may include the steps of:

In the first step, the execution subject may combine the target user feature information and the preprocessing scene information to obtain first combined information.

And a second step, the execution subject may input the first combination information into a second scene feature information generation model to generate second initial scene feature information. Wherein the second scene feature information generation model may be a model that generates scene feature information. In practice, the second scene characteristic information generation model may be a fully connected network.

And thirdly, the execution subject can perform information combination on the first initial scene feature information, the second initial scene feature information and the preprocessing scene information to obtain second combination information serving as scene feature information aiming at the preprocessing scene information.

Step 304, generating an information feature set according to the user feature information set and the scene feature information set.

In some embodiments, the executing entity may generate an information feature set according to the user feature information set and the scene feature information set. The information features in the information feature set are corresponding user feature information and scene feature information.

As an example, the execution subject may perform corresponding information combination of the user feature information in the user feature information set and the scene feature information in the scene feature information set, to obtain a feature combination information set as the information feature set.

In some optional implementations of some embodiments, the executing entity may input the information feature set to a classification model corresponding to the target classification task to generate a classification result. The target classification task may be a task for classification that is set in advance. In practice, the target classification task may be a classification task of personal sensitive information, and may also be a classification task of personal sensitive information. The model structure of the classification model is dynamically set according to the target classification task. The specific training method is not described here in detail.

With further reference to FIG. 4, a flow 400 of further embodiments of a sensitive information feature extraction method according to the present disclosure is shown. The sensitive information characteristic extraction method comprises the following steps:

Step 401, performing information preprocessing on each user sensitive information in the sample data set to generate preprocessed user information, so as to obtain a preprocessed user information set, and performing information preprocessing on each scene description information in the corresponding scene description information set to generate preprocessed scene information, so as to obtain a preprocessed scene information set.

And step 402, in response to determining that the mode form of the user sensitive information corresponding to the preprocessed user information is a text form, inputting the preprocessed user information into the bidirectional time sequence neural network model to generate hidden layer output information for each moment.

In some embodiments, in response to determining that the modality form of the user-sensitive information corresponding to the pre-processed user information is a text form, an executing subject (e.g., the electronic device 101 shown in fig. 1) may input the pre-processed user information to a bi-directional time-sequential neural network model to generate hidden layer output information for each moment. In practice, the two-way sequential neural network model may be a two-way Long Short-Term Memory network (Bi-LSTM).

Step 403, performing weighted summation processing on the obtained hidden layer output information by using the first attention mechanism model to generate attribute content characteristic information as user characteristic information.

In some embodiments, the executing body may perform weighted summation processing on the obtained hidden layer output information by using the first attention mechanism model to generate attribute content feature information as the user feature information. Wherein the first attention mechanism model may be a transducer model.

As an example, the executing entity may determine weight information corresponding to each hidden output information in the respective hidden output information using the first attention mechanism model. Then, each piece of hidden output information is multiplied with the corresponding weight information to generate a multiplication result, and a multiplication result set is obtained. And finally, adding the multiplication results in the multiplication result set to obtain attribute content characteristic information serving as user characteristic information.

And step 404, in response to determining that the mode form of the user sensitive information corresponding to the preprocessed user information is an image form, performing multidimensional convolution on the preprocessed user information to obtain a plurality of convolution results.

In some embodiments, in response to determining that the modality form of the user sensitive information corresponding to the preprocessed user information is an image form, the executing body may perform multidimensional convolution on the preprocessed user information to obtain a plurality of convolution results. The convolution result can represent convolution characteristic information corresponding to the sensitive information of the corresponding user.

As an example, the execution body may perform multidimensional convolution on the preprocessed user information using a plurality of different convolution kernels, to obtain a plurality of convolution results.

And step 405, splicing each convolution result in the plurality of convolution results to obtain a spliced result.

In some embodiments, the execution body may splice each convolution result of the plurality of convolution results to obtain a spliced result.

Step 406, inputting the splicing result into an image feature information generation model to generate attribute content feature information corresponding to the preprocessing attribute content information as user feature information.

In some embodiments, the execution body may input the stitching result to an image feature information generation model to generate attribute content feature information corresponding to the preprocessed attribute content information as the user feature information. The image feature information generation model may be a model that generates image feature information. In practice, the image characteristic information generation model may be a fully connected layer of a multi-layer serial connection.

Step 407, extracting feature information of each piece of preprocessing scene information in the preprocessing scene information set according to the user feature information set to generate scene feature information, thereby obtaining a scene feature information set.

Step 408, generating an information feature set according to the user feature information set and the scene feature information set.

In some embodiments, the specific implementation of steps 401, 407-408 and the technical effects thereof may refer to steps 301, 303-304 in the corresponding embodiment of fig. 3, which are not described herein.

As can be seen from fig. 4, compared to the description of some embodiments corresponding to fig. 2, the process 400 of the sensitive information feature extraction method in some embodiments corresponding to fig. 4 can perform feature extraction accurately for user sensitive information in a text form in a modality form and in an image form in a modality form.

With further reference to fig. 5, as an implementation of the method shown in the above figures, the present disclosure provides embodiments of a sensitive information feature extraction apparatus, which correspond to those method embodiments shown in fig. 2, and which are particularly applicable in various electronic devices.

As shown in fig. 5, a sensitive information feature extraction apparatus 500 includes a preprocessing unit 501, a first extraction unit 502, a second extraction unit 503, and a generation unit 504. The preprocessing unit 501 is configured to perform information preprocessing on each piece of user sensitive information in a sample data set to generate preprocessed user information to obtain a preprocessed user information set, and perform information preprocessing on each piece of scene description information in a corresponding scene description information set to generate preprocessed scene information to obtain a preprocessed scene information set, wherein the sample data set comprises user sensitive information in a multi-mode form, the first extracting unit 502 is configured to extract characteristic information of each piece of preprocessed user information in the preprocessed user information set to generate user characteristic information to obtain a user characteristic information set, the second extracting unit 503 is configured to extract characteristic information of each piece of preprocessed scene information in the preprocessed scene information set according to the user characteristic information set to generate scene characteristic information to obtain a scene characteristic information set, and the generating unit 504 is configured to generate an information characteristic set according to the user characteristic information set and the scene characteristic information set.

In some optional implementations of some embodiments, the user sensitive information in the sample data set includes user attribute information and user attribute content information, and the preprocessing unit 501 may be further configured to perform information preprocessing on the user attribute information included in the user sensitive information to obtain preprocessed user attribute information, and perform information preprocessing on the user attribute content information included in the user sensitive information to obtain preprocessed attribute content information, and generate preprocessed user information for the user sensitive information according to the preprocessed user attribute information and the preprocessed attribute content information.

In some optional implementations of some embodiments, the first extracting unit 502 may be further configured to extract, as the user feature information, attribute content feature information corresponding to the preprocessing attribute content information included in the preprocessing user information.

In some optional implementations of some embodiments, the second extraction unit 503 may be further configured to determine user feature information corresponding to the above-mentioned preprocessed scene information as target user feature information, input preprocessed user attribute information corresponding to the above-mentioned target user feature information and the above-mentioned preprocessed scene information into a first scene feature information generation model based on an attention mechanism to generate first initial scene feature information, and generate scene feature information for the above-mentioned preprocessed scene information according to the above-mentioned first initial scene feature information.

In some optional implementations of some embodiments, the second extracting unit 503 may be further configured to combine the target user feature information and the preprocessing scene information to obtain first combined information, input the first combined information to a second scene feature information generating model to generate second initial scene feature information, and combine the first initial scene feature information, the second initial scene feature information, and the preprocessing scene information to obtain second combined information as scene feature information for the preprocessing scene information.

In some optional implementations of some embodiments, the first extraction unit 502 may be further configured to, in response to determining that the modality form of the user sensitive information corresponding to the preprocessed user information is text form, input the preprocessed user information to the bidirectional time-sequential neural network model to generate hidden layer output information for each moment, and perform weighted summation processing on each obtained hidden layer output information by using the first attention mechanism model to generate attribute content feature information as the user feature information.

In some optional implementations of some embodiments, the first extracting unit 502 may be further configured to, in response to determining that the modality form of the user sensitive information corresponding to the preprocessed user information is an image form, perform multidimensional convolution on the preprocessed user information to obtain a plurality of convolution results, splice each convolution result of the plurality of convolution results to obtain a spliced result, and input the spliced result to an image feature information generating model to generate attribute content feature information corresponding to the preprocessed attribute content information as the user feature information.

In some optional implementations of some embodiments, the first extracting unit 502 may be further configured to, in response to determining that the modality form of the user sensitive information corresponding to the preprocessed user information is a video form, perform, for each frame of image in the frame image sequence corresponding to the preprocessed user information, a processing step of performing multidimensional convolution on the image to obtain a plurality of convolution results, stitching each convolution result of the plurality of convolution results to obtain a stitching result, inputting the stitching result to an image feature information generating model to generate attribute content feature information corresponding to the preprocessed attribute content information as target attribute content feature information, inputting the obtained target attribute content feature information sequence to a time-sequential neural network model to output a hidden layer output result set at each moment, and performing weighted summation processing on each hidden layer output information in the hidden layer output result set by using a first attention mechanism model to generate attribute content feature information as user feature information.

In some optional implementations of some embodiments, the first extraction unit 502 may be further configured to generate, as the user feature information, attribute content feature information for the pre-processed user information using an audio feature information generation model in response to determining that the modality form of the user sensitive information corresponding to the pre-processed user information is an audio form.

In some alternative implementations of some embodiments, the sensitive information feature extraction device 500 further includes an input unit. The input unit may be further configured to input the information feature set to a classification model corresponding to the target classification task to generate a classification result.

It will be appreciated that the elements described in the sensitive information feature extraction device 500 correspond to the various steps in the method described with reference to fig. 2. Thus, the operations, features and resulting benefits described above for the method are equally applicable to the sensitive information feature extraction device 500 and the units contained therein, and are not described in detail herein.

Referring now to fig. 6, a schematic diagram of an electronic device 600 (e.g., electronic device 101 of fig. 1) suitable for use in implementing some embodiments of the present disclosure is shown. The electronic device shown in fig. 6 is merely an example and should not impose any limitations on the functionality and scope of use of embodiments of the present disclosure.

As shown in fig. 6, the electronic device 600 may include a processing means (e.g., a central processing unit, a graphics processor, etc.) 601, which may perform various appropriate actions and processes according to programs stored in a read-only memory 602 or programs loaded from a storage 608 into a random access memory 603. In the random access memory 603, various programs and data necessary for the operation of the electronic device 600 are also stored. The processing means 601, the read only memory 602 and the random access memory 603 are connected to each other via a bus 604. An input/output interface 605 is also connected to the bus 604.

In general, devices may be connected to the input/output interface 605 including input devices 606 such as a touch screen, touch pad, keyboard, mouse, camera, microphone, accelerometer, gyroscope, etc., output devices 607 including a Liquid Crystal Display (LCD), speaker, vibrator, etc., storage devices 608 including magnetic tape, hard disk, etc., and communication devices 609. The communication means 609 may allow the electronic device 600 to communicate with other devices wirelessly or by wire to exchange data. While fig. 6 shows an electronic device 600 having various means, it is to be understood that not all of the illustrated means are required to be implemented or provided. More or fewer devices may be implemented or provided instead. Each block shown in fig. 6 may represent one device or a plurality of devices as needed.

In particular, according to some embodiments of the present disclosure, the processes described above with reference to flowcharts may be implemented as computer software programs. For example, some embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method shown in the flow chart. In such embodiments, the computer program may be downloaded and installed from a network via communications device 609, or from storage device 608, or from read only memory 602. The above-described functions defined in the methods of some embodiments of the present disclosure are performed when the computer program is executed by the processing device 601.

It should be noted that, in some embodiments of the present disclosure, the computer readable medium may be a computer readable signal medium or a computer readable storage medium, or any combination of the two. The computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples of a computer-readable storage medium may include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In some embodiments of the present disclosure, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In some embodiments of the present disclosure, however, the computer-readable signal medium may comprise a data signal propagated in baseband or as part of a carrier wave, with the computer-readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to electrical wiring, fiber optic cable, RF (radio frequency), and the like, or any suitable combination of the foregoing.

In some embodiments, the clients, servers may communicate using any currently known or future developed network protocol, such as HTTP (HyperText Transfer Protocol ), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), the internet (e.g., the internet), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future developed networks.

The computer readable medium may be included in the electronic device or may exist alone without being incorporated into the electronic device. The computer readable medium carries one or more programs which, when executed by the electronic device, cause the electronic device to perform information preprocessing on each user-sensitive information in a sample dataset to generate preprocessed user information sets, and perform information preprocessing on each scene description information in a corresponding scene description information set to generate preprocessed scene information sets, wherein the sample dataset comprises multi-modal user-sensitive information, extract feature information of each preprocessed user information in the preprocessed user information sets to generate user feature information to obtain a user feature information set, extract feature information of each preprocessed scene information in the preprocessed scene information sets according to the user feature information sets to generate scene feature information to obtain scene feature information sets, and generate information feature sets according to the user feature information sets and the scene feature information sets.

Computer program code for carrying out operations for some embodiments of the present disclosure may be written in one or more programming languages, including an object oriented programming language such as Java, smalltalk, C ++ and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computer (for example, through the Internet using an Internet service provider).

The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

The units described in some embodiments of the present disclosure may be implemented by means of software, or may be implemented by means of hardware. The described units may also be provided in a processor, which may be described, for example, as a processor comprising a preprocessing unit, a first extraction unit, a second extraction unit and a generation unit. The names of these units do not constitute a limitation on the unit itself in some cases, and the generating unit may also be described as "a unit that generates an information feature set from the above-described user feature information set and the above-described scene feature information set", for example.

The functions described above herein may be performed, at least in part, by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic that may be used include Field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), systems-on-a-chip (SOCs), complex Programmable Logic Devices (CPLDs), and the like.

Some embodiments of the present disclosure also provide a computer program product comprising a computer program which, when executed by a processor, implements any of the sensitive information feature extraction methods described above.

The foregoing description is only of the preferred embodiments of the present disclosure and description of the principles of the technology being employed. It will be appreciated by those skilled in the art that the scope of the invention in the embodiments of the present disclosure is not limited to the specific combination of the above technical features, but encompasses other technical features formed by any combination of the above technical features or their equivalents without departing from the spirit of the invention. Such as the above-described features, are mutually substituted with (but not limited to) the features having similar functions disclosed in the embodiments of the present disclosure.

Claims

1. A sensitive information feature extraction method comprises the following steps:

Performing information preprocessing on each user sensitive information in a sample data set to generate preprocessed user information, obtaining preprocessed user information sets, performing information preprocessing on each scene description information in corresponding scene description information sets to generate preprocessed scene information, obtaining preprocessed scene information sets, wherein the sample data sets comprise user sensitive information in a multi-mode form, the user sensitive information in the sample data sets comprises user attribute information and user attribute content information, the information preprocessing is performed on each user sensitive information in the sample data sets to generate preprocessed user information, and the information preprocessing is performed on the user attribute information included in the user sensitive information to obtain preprocessed user attribute information, and the information preprocessing is performed on the user attribute content information included in the user sensitive information to obtain preprocessed attribute content information;

Extracting the characteristic information of each piece of preprocessing user information in the preprocessing user information set to generate user characteristic information to obtain a user characteristic information set, wherein the extracting the characteristic information of each piece of preprocessing user information in the preprocessing user information set to generate user characteristic information comprises the steps of extracting attribute content characteristic information corresponding to preprocessing attribute content information included in the preprocessing user information as user characteristic information;

Extracting feature information of each piece of preprocessing scene information in the preprocessing scene information set according to the user feature information set to generate scene feature information, and obtaining a scene feature information set;

And generating an information feature set according to the user feature information set and the scene feature information set.

2. The method of claim 1, wherein the extracting feature information of each of the preprocessed scene information sets from the set of user feature information to generate scene feature information comprises:

Determining user characteristic information corresponding to the preprocessing scene information as target user characteristic information;

Inputting the preprocessed user attribute information and the preprocessed scene information corresponding to the target user feature information into a first scene feature information generation model based on an attention mechanism to generate first initial scene feature information;

and generating scene characteristic information aiming at the preprocessing scene information according to the first initial scene characteristic information.

3. The method of claim 2, wherein the generating scene feature information for the pre-processing scene information from the first initial scene feature information comprises:

Combining the target user characteristic information and the preprocessing scene information to obtain first combined information;

inputting the first combined information into a second scene feature information generation model to generate second initial scene feature information;

And combining the first initial scene characteristic information, the second initial scene characteristic information and the preprocessing scene information to obtain second combined information serving as scene characteristic information aiming at the preprocessing scene information.

4. The method of claim 1, wherein the extracting attribute content feature information corresponding to the preprocessing attribute content information included in the preprocessing user information as the user feature information includes:

In response to determining that the mode form of the user sensitive information corresponding to the preprocessed user information is a text form, inputting the preprocessed user information into a bidirectional time sequence neural network model to generate hidden layer output information for each moment;

and carrying out weighted summation processing on the obtained hidden layer output information by using the first attention mechanism model to generate attribute content characteristic information serving as user characteristic information.

5. The method of claim 1, wherein the extracting attribute content feature information corresponding to the preprocessing attribute content information included in the preprocessing user information as the user feature information includes:

Responding to the fact that the mode form of the user sensitive information corresponding to the preprocessed user information is an image form, and carrying out multidimensional convolution on the preprocessed user information to obtain a plurality of convolution results;

Splicing all convolution results in the plurality of convolution results to obtain a splicing result;

And inputting the splicing result into an image characteristic information generation model to generate attribute content characteristic information corresponding to the preprocessing attribute content information as user characteristic information.

6. The method of claim 1, wherein the extracting attribute content feature information corresponding to the preprocessing attribute content information included in the preprocessing user information as the user feature information includes:

In response to determining that the modality form of the user sensitive information corresponding to the pre-processed user information is a video form, for each frame of image in the sequence of frame images corresponding to the pre-processed user information, performing the following processing steps:

Carrying out multidimensional convolution on the image to obtain a plurality of convolution results;

inputting the splicing result into an image characteristic information generation model to generate attribute content characteristic information corresponding to the preprocessing attribute content information as target attribute content characteristic information;

Inputting the obtained target attribute content characteristic information sequence into a time sequence neural network model to output hidden layer output result sets at all times;

And carrying out weighted summation processing on each hidden layer output information in the hidden layer output result set by using a first attention mechanism model so as to generate attribute content characteristic information serving as user characteristic information.

7. The method of claim 1, wherein the extracting attribute content feature information corresponding to the preprocessing attribute content information included in the preprocessing user information as the user feature information includes:

and generating attribute content characteristic information aiming at the preprocessing user information as the user characteristic information by utilizing an audio characteristic information generation model in response to determining that the mode form of the user sensitive information corresponding to the preprocessing user information is an audio form.

8. The method of claim 1, wherein the method further comprises:

and inputting the information feature set into a classification model corresponding to the target classification task to generate a classification result.

9. A sensitive information feature extraction apparatus comprising:

The preprocessing unit is configured to perform information preprocessing on each user sensitive information in a sample data set to generate preprocessed user information to obtain preprocessed user information sets, and perform information preprocessing on each scene description information in the corresponding scene description information set to generate preprocessed scene information to obtain preprocessed scene information sets, wherein the sample data set comprises user sensitive information in a multi-mode form, the user sensitive information in the sample data set comprises user attribute information and user attribute content information, the information preprocessing is performed on each user sensitive information in the sample data set to generate preprocessed user information, and the information preprocessing is performed on the user attribute information included in the user sensitive information to obtain preprocessed user attribute information, and the information preprocessing is performed on the user attribute content information included in the user sensitive information to obtain preprocessed attribute content information;

A first extraction unit configured to extract feature information of each piece of preprocessed user information in the preprocessed user information set to generate user feature information, thereby obtaining a user feature information set;

The second extracting unit is configured to extract feature information of each piece of preprocessing scene information in the preprocessing scene information set according to the user feature information set to generate scene feature information to obtain the scene feature information set, wherein the extracting of the feature information of each piece of preprocessing user information in the preprocessing user information set to generate the user feature information comprises the steps of extracting attribute content feature information corresponding to preprocessing attribute content information included in the preprocessing user information as user feature information;

And the generating unit is configured to generate an information feature set according to the user feature information set and the scene feature information set.

10. An electronic device, comprising:

one or more processors;

a storage device having one or more programs stored thereon,

When executed by the one or more processors, causes the one or more processors to implement the method of any of claims 1-8.

11. A computer readable medium having stored thereon a computer program, wherein the computer program, when executed by a processor, implements the method of any of claims 1-8.

12. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any of claims 1-8.