Dialogue stream-based leader pretending short message fraud identification method and system

Abstract

The embodiment of the invention discloses a conversation stream-based leader pretending short message fraud identification method and system. The method comprises the following steps of: obtaining multiple pieces of short message data; merging the multiple pieces of short message data into a short message text, and calculating a weight matrix corresponding to the short message text; and obtaining a short message recognition result through the Bi-LSTM model according to the short message text and the weight matrix corresponding to the short message text. By implementing the embodiment of the invention, the accuracy of short message identification can be improved.

Description

Technical Field

[0001] The present application relates to the field of big data technology, and in particular to a method and system for identifying SMS fraud impersonating leaders based on dialogue flow. Background Art

[0002] Criminals pretend to be leaders to gain trust by showing concern for work and life, and then ask victims to transfer money to a designated account on the grounds that relatives, friends or leaders are in urgent need of funds and it is inconvenient for them to transfer money directly, thus committing fraud. Special characters / pictographs / phono-semantic characters and other variant characters are used extensively in port-type SMS, 5G messages, and MMS to circumvent existing prevention strategies, and most current SMS fraud identification methods rely on content analysis of a single SMS. This method has limitations because a single SMS may not be enough to fully reflect the intention and background of the conversation. Fraudulent SMS impersonating leaders often lure victims step by step through multiple SMS messages, so a more comprehensive method is needed to identify fraudulent SMS. Summary of the invention

[0003] The embodiment of the present application discloses a method and system for identifying SMS fraud by impersonating a leader based on dialogue flow, which can improve the accuracy of SMS recognition.

[0004] The present application embodiment discloses a method for identifying SMS fraud impersonating a leader based on a dialogue flow, the method comprising: Get multiple SMS data; Merge the multiple SMS data into SMS text, and calculate the weight matrix corresponding to the SMS text; The Bi-LSTM model is used to obtain a text message recognition result based on the text message and a weight matrix corresponding to the text message.

[0005] As an optional implementation manner, the calculating the weight matrix corresponding to the SMS text includes: Performing word segmentation processing on the SMS text to obtain multiple SMS word segments; Calculate the word frequency and inverse document frequency of each SMS word segment; The weight matrix corresponding to the SMS text is obtained according to the product of the word frequency and the inverse document frequency of each SMS word segment.

[0006] As an optional implementation manner, after obtaining the plurality of SMS data, the method further includes: Performing data cleaning on the plurality of SMS data to obtain a plurality of cleaned SMS data; The step of merging the plurality of SMS data into SMS text and calculating a weight matrix corresponding to the SMS text includes: The multiple cleaned SMS data are merged into SMS text, and a weight matrix corresponding to the SMS text is calculated.

[0007] As an optional implementation, the Bi-LSTM model includes a word vector input layer; the obtaining of the SMS recognition result by the Bi-LSTM model according to the SMS text and a weight matrix corresponding to the SMS text includes: Perform word segmentation processing on the SMS text through a word vector input layer to obtain multiple SMS word vectors; Performing feature extraction on the multiple SMS word vectors to obtain SMS feature vectors; The SMS recognition result is obtained according to the SMS feature vector and the weight matrix corresponding to the SMS text.

[0008] As an optional implementation manner, the extracting features of the multiple SMS word vectors to obtain SMS feature vectors includes: Simultaneously, feature extraction is performed on the multiple SMS word vectors from a forward direction and a backward direction to obtain the SMS feature vectors.

[0009] As an optional implementation manner, obtaining the SMS recognition result according to the SMS feature vector and the weight matrix corresponding to the SMS text includes: According to the weight matrix corresponding to the text message, the text message feature vector is weighted to obtain a weighted text message feature vector; The weighted SMS feature vector is classified to obtain the SMS recognition result.

[0010] As an optional implementation, after obtaining the SMS recognition result according to the SMS text and the weight matrix corresponding to the SMS text through the Bi-LSTM model, the method further includes: If the message identification result indicates that the message is an abnormal message, the message text is recorded and the number corresponding to the message data is shut down.

[0011] The present application embodiment discloses a system for identifying SMS fraud impersonating a leader based on a dialogue flow, the system comprising: The acquisition module is used to obtain multiple SMS data; A merging module, used to merge the multiple SMS data into SMS text, and calculate the weight matrix corresponding to the SMS text; The recognition module is used to obtain a text message recognition result according to the text message and a weight matrix corresponding to the text message through a Bi-LSTM model.

[0012] An embodiment of the present application discloses an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor implements any one of the methods for identifying SMS fraud impersonating a leader based on a dialogue flow disclosed in the embodiment of the present application.

[0013] An embodiment of the present application discloses a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, any one of the methods for identifying SMS fraud impersonating a leader based on a dialogue flow disclosed in the embodiment of the present application is implemented.

[0014] Compared with the related art, the embodiments of the present application have the following beneficial effects: The embodiment of the present application provides a method and system for identifying SMS fraud impersonating leaders based on dialogue flow, obtaining multiple SMS data; merging multiple SMS data into SMS text, and calculating the weight matrix corresponding to the SMS text; obtaining the SMS recognition result according to the SMS text and the weight matrix corresponding to the SMS text through the Bi-LSTM model. In implementing the embodiment of the present application, by merging multiple SMS data into SMS text and calculating the weight matrix corresponding to the SMS text, it is avoided to obtain the SMS recognition result through a small amount of SMS data, thereby affecting the accuracy of the SMS recognition result. At the same time, the weight matrix corresponding to the SMS text is calculated, and the SMS recognition result is obtained according to the SMS text and the weight matrix corresponding to the SMS text through the Bi-LSTM model. The words that need to be paid attention to in the SMS text are highlighted according to the weight matrix corresponding to the SMS text, which can improve the accuracy of SMS recognition while enhancing the robustness of the Bi-LSTM model. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 It is a flow chart of a method for identifying SMS fraud impersonating a leader based on a dialogue flow disclosed in an embodiment of the present application; Figure 2 It is a flowchart of obtaining a text message recognition result according to a text message and a weight matrix corresponding to the text message by using a Bi-LSTM model disclosed in an embodiment of the present application; Figure 3 It is a structural diagram of a system for identifying SMS fraud impersonating leaders based on dialogue flow disclosed in an embodiment of the present application. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0018] It should be noted that the terms "including" and "having" and any variations thereof in the embodiments of the present application and the accompanying drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device including a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices.

[0019] The embodiment of the present application discloses a method and system for identifying SMS fraud impersonating a leader based on a dialogue flow, which can improve the accuracy of SMS recognition. Detailed descriptions are given below.

[0020] Figure 1 This is a flow chart of a method for identifying SMS fraud impersonating a leader based on a dialogue flow disclosed in an embodiment of the present application. Figure 1 The described method for identifying SMS fraud impersonating a leader based on a conversation flow is applicable to electronic devices, which may include but are not limited to mobile phones, tablet computers, wearable devices, laptop computers, PCs (Personal Computers), etc., and the present application embodiment does not limit this. Figure 1 As shown, the method for identifying SMS fraud impersonating a leader based on a dialogue flow may include the following steps: Step S101, obtaining multiple SMS data.

[0021] In some embodiments, an electronic device may collect multiple SMS data generated by each of the electronic devices, or multiple SMS data generated by mobile phones and other electronic devices, so that the electronic device obtains multiple SMS data generated in the same electronic device. Among them, each SMS data may include SMS content, SMS sender, SMS receiver and SMS timestamp. The SMS content can be used to record the text content presented in the SMS data; the SMS sender can be used to indicate the device identification of the electronic device that sent the SMS data. If the electronic device is a mobile phone or other device that can communicate, the device identification can refer to the mobile phone number; the SMS receiver can be used to indicate the device identification of the electronic device that receives the SMS data; the SMS timestamp can be used to record the time point when the SMS data is received. For example, as shown in Table 1, Table 1 can represent a piece of SMS data: Among them, the SMS sender in the SMS data can be "159XXXXXX26", the SMS timestamp can be "August 4, 2024 14:40", the SMS receiver can be "179XXXXXXXX", and the SMS content can be "Hello".

[0022] Step S102, merging multiple SMS data into SMS text, and calculating the weight matrix corresponding to the SMS text.

[0023] In some embodiments, after obtaining multiple SMS data, the electronic device may merge the multiple SMS data into SMS text. Further, the electronic device may merge the SMS contents in the multiple SMS data to obtain the SMS text. Further, the electronic device may merge the SMS contents in the multiple SMS data according to the SMS timestamps corresponding to the multiple SMS data, thereby obtaining the SMS text. Specifically, the SMS timestamps corresponding to the multiple SMS data may be merged from morning to night. For example, if SMS data A includes SMS content A and SMS timestamp 14:40 on August 4, 2024, SMS data B includes SMS content B and SMS timestamp 14:41 on August 4, 2024, and SMS data C includes SMS content C and SMS timestamp 14:42 on August 4, 2024, the obtained SMS text may be [SMS content A, SMS content B, SMS content C].

[0024] In some embodiments, the electronic device may perform data cleaning on multiple SMS data to obtain multiple cleaned SMS data; merge the multiple cleaned SMS data into SMS text, and calculate the weight matrix corresponding to the SMS text. Among them, data cleaning on multiple SMS data may include, for the SMS content of the multiple SMS data, using the Python toolkit langconv to convert Chinese traditional and simplified characters, and using pypinyin to convert pinyin, identifying homophones, using regular expressions to identify and replace special characters, including variant characters such as pictographic characters and phono-semantic characters, cleaning special characters and blank characters, and building a custom mapping table to identify and convert SMS content, etc.

[0025] The electronic device performs data cleaning on multiple SMS data, obtains multiple cleaned SMS data, merges the multiple cleaned SMS data into SMS text, and calculates the weight matrix corresponding to the SMS text. This can reduce the content of the SMS containing one or more of the variant characters such as special characters, pictographic characters, and phono-semantic characters, thereby affecting the judgment of the Bi-LSTM model on the SMS text. The purpose is to regularize the SMS content and enable the Bi-LSTM model to better identify fraudulent behavior.

[0026] As an optional implementation, the electronic device performs word segmentation processing on the SMS text to obtain multiple SMS word segmentations; calculates the word frequency and inverse document frequency of each SMS word segmentation; and obtains the weight matrix corresponding to the SMS text according to the product of the word frequency and inverse document frequency of each SMS word segmentation. The electronic device can use HanLP to perform tasks such as part-of-speech tagging, named entity recognition, and dependency syntax analysis to help understand the basic grammatical structure and vocabulary importance of the SMS text. For example, if the SMS text is "I love China", then after the SMS text is segmented, the SMS word segmentations "I", "love" and "China" can be obtained. The word frequency of the SMS word segmentation can represent the frequency of occurrence of the SMS word segmentation. The calculation of the word frequency of the SMS word segmentation can be shown in the following formula: in, It can represent the word frequency of SMS segmentation. It can indicate the number of times the SMS word appears in the SMS text. It can represent the total number of occurrences of all SMS words in the SMS text.

[0027] The calculation of the inverse document frequency of SMS word segmentation can be shown as follows: in, represents the inverse document frequency of the SMS word segmentation, represents the total number of documents in the corpus for training the module that computes the weight matrix, Indicates the total number of documents in the corpus that contain the SMS token.

[0028] The electronic device obtains the weight matrix corresponding to the SMS text according to the product of the word frequency and the inverse document frequency of each SMS word segment, wherein the weight matrix corresponding to the SMS text can be in, Represents the weight matrix corresponding to the SMS text, Represents the product of the word frequency and inverse document frequency of each SMS word.

[0029] The electronic device performs word segmentation processing on the SMS text to obtain multiple SMS segmentations; calculates the word frequency and inverse document frequency of each SMS segmentation, and obtains the weight matrix corresponding to the SMS text according to the product of the word frequency and inverse document frequency of each SMS segmentation. It can identify important words, and use the weight matrix to represent the SMS text, highlighting those words that frequently appear in the conversation but are less common in the entire corpus, which can improve the accuracy of subsequent SMS recognition through the Bi-LSTM model.

[0030] Step S103, obtaining a text message recognition result through a Bi-LSTM model according to the text message and a weight matrix corresponding to the text message.

[0031] In some embodiments, the electronic device can fuse the SMS text and the weight matrix corresponding to the SMS text through the Bi-LSTM model to obtain the SMS recognition result. Among them, the SMS recognition result can include two types, namely "normal SMS" and "abnormal SMS". If the SMS recognition result is "normal SMS", it means that the current SMS data does not contain fraudulent content; if the SMS recognition result is "abnormal SMS", it means that the current SMS data contains fraudulent content.

[0032] As an optional implementation, the electronic device can train the Bi-LSTM model to be trained through a sample SMS text set to obtain a Bi-LSTM model that has been trained. The sample SMS text set may include multiple sample SMS texts and a weight matrix and a target recognition result corresponding to each sample SMS text. The electronic device inputs the sample SMS text set into the Bi-LSTM model to be trained, and calculates the predicted recognition result corresponding to each sample SMS text through the Bi-LSTM model to be trained according to each sample SMS text and the weight matrix corresponding to each sample SMS text. According to the error between the target recognition result and the predicted recognition result corresponding to each sample SMS text, the model data of the Bi-LSTM model to be trained is adjusted until the error between the target recognition result and the predicted recognition result corresponding to each sample SMS text is less than a preset error threshold, so as to obtain a Bi-LSTM model that has been trained.

[0033] In some embodiments, after obtaining the SMS recognition result through the Bi-LSTM model according to the SMS text and the weight matrix corresponding to the SMS text, if the SMS recognition result indicates an abnormal SMS, the SMS text is recorded and the number corresponding to the SMS data is shut down. After using the Bi-LSTM model for identification, the number identified as an abnormal SMS number is output and shut down, and the fraudulent speech is recorded as a positive sample, and the number with the same speech is traced back and shut down. If the SMS recognition result indicates a normal SMS, the SMS data of the next time period is re-acquired as new SMS data, and multiple new SMS data are merged into SMS text, and the weight matrix corresponding to the SMS text is calculated. The SMS recognition result is obtained through the Bi-LSTM model according to the SMS text and the weight matrix corresponding to the SMS text, so as to monitor the SMS data in real time to improve the security and anti-fraud capabilities of electronic equipment.

[0034] In an embodiment of the present application, an electronic device obtains multiple SMS data, merges the multiple SMS data into SMS text, and calculates the weight matrix corresponding to the SMS text, and obtains the SMS recognition result according to the SMS text and the weight matrix corresponding to the SMS text through the Bi-LSTM model. By merging multiple SMS data into SMS text and calculating the weight matrix corresponding to the SMS text, it is avoided to obtain the SMS recognition result through a small amount of SMS data, thereby affecting the accuracy of the SMS recognition result. At the same time, the weight matrix corresponding to the SMS text is calculated, and the SMS recognition result is obtained according to the SMS text and the weight matrix corresponding to the SMS text through the Bi-LSTM model. According to the weight matrix corresponding to the SMS text, the words that need to be paid attention to in the SMS text are highlighted, while enhancing the robustness of the Bi-LSTM model, the accuracy of SMS recognition can be improved.

[0035] Figure 2 1 is a flow chart of obtaining a text message recognition result according to a text message and a weight matrix corresponding to the text message by using a Bi-LSTM model disclosed in an embodiment of the present application. Figure 2 As shown, the steps are to obtain the SMS recognition result according to the SMS text and the weight matrix corresponding to the SMS text through the Bi-LSTM model, and also include the following steps: Step S201, performing word segmentation processing on the SMS text through the word vector input layer to obtain multiple SMS word vectors.

[0036] In some embodiments, the Bi-LSTM model may include a word vector input layer, a network layer, a weighted self-attention layer, and a feature classification layer. Among them, the word vector input layer can perform word segmentation on the SMS text according to a word segmentation tool (such as HanLP), and convert the segmented SMS text into SMS word vectors, thereby obtaining multiple SMS word vectors. The network layer can be used to extract features from the SMS word vectors to obtain SMS feature vectors. The weighted self-attention layer can be used to weight the extracted SMS feature vectors in combination with the weight matrix, and highlight important words through the self-attention mechanism to obtain weighted SMS feature vectors. The feature classification layer is used to classify the weighted SMS feature vectors to obtain SMS recognition features.

[0037] As an optional implementation, the electronic device may perform word segmentation processing on the SMS text through a word vector input layer to obtain multiple SMS word segments, and convert each SMS word segment into a corresponding SMS word vector.

[0038] Step S202, extracting features from multiple SMS word vectors to obtain SMS feature vectors.

[0039] In some embodiments, the electronic device can use a Bi-LSTM network to extract SMS word vectors and retain context information. Furthermore, the electronic device can simultaneously perform feature extraction on multiple SMS word vectors from the forward and backward directions to obtain SMS feature vectors. For example, after the electronic device performs word segmentation processing on the SMS text "I love China" through the word vector input layer, it can obtain the SMS word vector A corresponding to the SMS segmentation "I", the SMS word vector B corresponding to the SMS segmentation "love", and the SMS word vector C corresponding to the SMS segmentation "China". The SMS word vector A, the SMS word vector B, and the SMS word vector C are input into the Bi-LSTM network for feature extraction. The forward network extracts features from the three word vectors in turn to obtain the first feature vector group. The backward network extracts features from the three word vectors in turn to obtain the second feature vector group Finally, the first feature vector group and the second feature vector group are concatenated to obtain the SMS feature vector corresponding to the SMS text. .

[0040] The electronic device simultaneously extracts features from multiple SMS word vectors from both forward and backward directions to obtain SMS feature vectors, which can capture contextual information from both forward and backward directions to better understand the semantics of the sentence.

[0041] Step S203, obtaining a text message recognition result according to the text message feature vector and the weight matrix corresponding to the text message.

[0042] In some embodiments, the electronic device may weight the SMS feature vector according to the weight matrix corresponding to the SMS text to obtain the weighted SMS feature vector, and classify the weighted SMS feature vector to obtain the SMS recognition result. Specifically, the electronic device may weight the SMS feature vector according to the weight matrix corresponding to the SMS text to obtain the weighted SMS feature vector; classify the weighted SMS feature vector to obtain the SMS recognition result.

[0043] Furthermore, the weighted self-attention layer may include a weighting module and a self-attention module. The electronic device may weight the SMS feature vector according to the weight matrix through the weighting module in the weighted self-attention layer to obtain the weighted SMS feature vector, and assign different weights to each SMS feature vector through the self-attention module in the weighted self-attention layer, thereby strengthening the proportion of key information, wherein the model parameters used for the specific weights assigned to each SMS feature vector can be obtained by training the Bi-LSTM model. When using the Bi-LSTM model to recognize SMS text, the self-attention module can automatically update the parameters from the information given by itself to assign different weights to different information. Specifically, the calculation formula of the weighted self-attention layer is as follows: , ， , , in, is the SMS feature vector, It can be expressed as , is the weight matrix corresponding to the SMS text, for dimensional identity matrix, is the length of the SMS text, is the weight matrix that is randomly initialized and learned during model training, is the weight matrix corresponding to the SMS text Output values ​​in the self-attention module.

[0044] Electronic devices will Input to the feature classification layer in the Bi-LSTM model, the activation function in the feature classification layer By classifying the weighted SMS feature vectors, the SMS recognition result can be obtained.

[0045] As an optional implementation method, after training the Bi-LSTM model, the performance of the Bi-LSTM model can be evaluated using indicators such as accuracy, recall, and F1 score. At the same time, the staff can also analyze cases of model prediction errors, find the shortcomings of the Bi-LSTM model, and adjust the model structure or retrain based on the results of the error analysis to improve the accuracy of SMS text recognition through the Bi-LSTM model.

[0046] In an embodiment of the present application, the electronic device performs word segmentation processing on the SMS text through a word vector input layer to obtain multiple SMS word vectors, performs feature extraction on the multiple SMS word vectors to obtain an SMS feature vector, and obtains an SMS recognition result based on the SMS feature vector and the weight matrix corresponding to the SMS text. It can perform feature extraction on the SMS text after word segmentation, and perform feature extraction on multiple SMS word vectors to obtain an SMS feature vector, which not only retains the text information of the SMS text, but also highlights important vocabulary that needs to be highlighted in the SMS text. Finally, based on the SMS feature vector and the weight matrix corresponding to the SMS text, the SMS recognition result is obtained, which can improve the accuracy of SMS recognition.

[0047] See also Figure 3 , Figure 31 is a schematic diagram of a system for identifying SMS fraud impersonating a leader based on a conversation flow disclosed in an embodiment of the present application. The system can be applied to the above-mentioned electronic device. Figure 3 As shown, a system 300 for identifying SMS fraud impersonating a leader based on a dialogue flow may include: an acquisition module 301, a merging module 302 and an identification module 303.

[0048] The acquisition module 301 is used to acquire multiple SMS data; A merging module 302 is used to merge multiple SMS data into SMS text and calculate the weight matrix corresponding to the SMS text; The recognition module 303 is used to obtain a text message recognition result according to the text message and the weight matrix corresponding to the text message by using the Bi-LSTM model.

[0049] In one embodiment, the merging module 302 is further used to perform word segmentation processing on the SMS text to obtain multiple SMS word segments; calculate the word frequency and inverse document frequency of each SMS word segment; and obtain a weight matrix corresponding to the SMS text according to the product of the word frequency and inverse document frequency of each SMS word segment.

[0050] In one embodiment, a system 300 for identifying SMS fraud impersonating a leader based on a dialog flow further includes a cleaning module: A cleaning module is used to clean the multiple SMS data to obtain multiple cleaned SMS data; The merging module 302 is further used to merge the multiple cleaned SMS data into SMS texts, and calculate the weight matrix corresponding to the SMS texts.

[0051] In one embodiment, the recognition module 303 further includes a word segmentation unit, a feature extraction unit, and a result recognition unit: The word segmentation unit is used to perform word segmentation processing on the SMS text through the word vector input layer to obtain multiple SMS word vectors; A feature extraction unit, used for extracting features from multiple SMS word vectors to obtain SMS feature vectors; The result recognition unit is used to obtain the SMS recognition result according to the SMS feature vector and the weight matrix corresponding to the SMS text.

[0052] In one embodiment, the feature extraction unit is also used to extract features from multiple SMS word vectors from both forward and backward directions to obtain SMS feature vectors.

[0053] In one embodiment, the result recognition unit is further used to perform weighted processing on the SMS feature vector according to the weight matrix corresponding to the SMS text to obtain a weighted SMS feature vector; and classify the weighted SMS feature vector to obtain an SMS recognition result.

[0054] In one embodiment, a system 300 for identifying SMS fraud by impersonating a leader based on a conversation flow further includes a recording module: The recording module is used to record the text of the SMS and shut down the number corresponding to the SMS data if the SMS identification result indicates that the SMS is an abnormal SMS.

[0055] An embodiment of the present application discloses a computer-readable storage medium storing a computer program, wherein when the computer program is executed by the processor, the processor implements any one of the methods for identifying SMS fraud impersonating a leader based on dialogue flow disclosed in the embodiment of the present application.

[0056] An embodiment of the present application discloses a computer program product, including a computer program, and the computer program can be executed by a processor to implement a method for identifying SMS fraud impersonating a leader based on dialogue flow as described in the above embodiments.

[0057] It should be understood that the "one embodiment" or "an embodiment" mentioned throughout the specification means that the specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. Those skilled in the art should also be aware that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required for the present application.

[0058] In the various embodiments of the present application, it should be understood that the size of the serial numbers of the above-mentioned processes does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0059] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0060] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0061] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-accessible memory. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product, which is stored in a memory and includes several requests for a computer device (which can be a personal computer, server or network device, etc., specifically a processor in a computer device) to perform some or all of the steps of the above-mentioned methods of various embodiments of the present application.

[0062] A person skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable programmable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0063] The above is a detailed introduction to a method and system for identifying SMS fraud impersonating a leader based on a dialogue flow disclosed in the embodiment of the present application. This article uses specific examples to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for those of ordinary skill in the art, according to the idea of ​​the present application, there will be changes in the specific implementation method and scope of application. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for identifying SMS fraud impersonating a leader based on dialogue flow, characterized in that: The method comprises: Get multiple SMS data; Merge the multiple SMS data into SMS text, and calculate the weight matrix corresponding to the SMS text; The Bi-LSTM model is used to obtain a text message recognition result based on the text message and a weight matrix corresponding to the text message.

2. According to the method of claim 1, the method is characterized by: The calculating the weight matrix corresponding to the SMS text includes: Performing word segmentation processing on the SMS text to obtain multiple SMS word segments; Calculate the word frequency and inverse document frequency of each SMS word segment; The weight matrix corresponding to the SMS text is obtained according to the product of the word frequency and the inverse document frequency of each SMS word segment.

3. According to the method of identifying SMS fraud impersonating a leader based on dialogue flow in claim 1, it is characterized in that: After obtaining the plurality of SMS data, the method further includes: Performing data cleaning on the plurality of SMS data to obtain a plurality of cleaned SMS data; The step of merging the plurality of SMS data into SMS text and calculating a weight matrix corresponding to the SMS text includes: The multiple cleaned SMS data are merged into SMS text, and a weight matrix corresponding to the SMS text is calculated.

4. According to the method of identifying SMS fraud impersonating a leader based on dialogue flow in claim 1, it is characterized in that: The Bi-LSTM model includes a word vector input layer; The obtaining of the SMS recognition result according to the SMS text and the weight matrix corresponding to the SMS text by the Bi-LSTM model includes: Perform word segmentation processing on the SMS text through a word vector input layer to obtain multiple SMS word vectors; Performing feature extraction on the multiple SMS word vectors to obtain SMS feature vectors; The SMS recognition result is obtained according to the SMS feature vector and the weight matrix corresponding to the SMS text.

5. According to the method of identifying SMS fraud impersonating a leader based on dialogue flow in claim 4, it is characterized in that: The extracting features of the multiple SMS word vectors to obtain SMS feature vectors includes: Simultaneously, feature extraction is performed on the multiple SMS word vectors from a forward direction and a backward direction to obtain the SMS feature vectors.

6. According to the method of identifying SMS fraud impersonating a leader based on dialogue flow in claim 4, it is characterized in that: The obtaining of the SMS recognition result according to the SMS feature vector and the weight matrix corresponding to the SMS text includes: According to the weight matrix corresponding to the text message, the text message feature vector is weighted to obtain a weighted text message feature vector; The weighted SMS feature vector is classified to obtain the SMS recognition result.

7. According to the method of claim 1, the method is characterized by: After obtaining the SMS recognition result according to the SMS text and the weight matrix corresponding to the SMS text by using the Bi-LSTM model, the method further includes: If the message identification result indicates that the message is an abnormal message, the message text is recorded and the number corresponding to the message data is shut down.

8. A system for identifying SMS fraud by impersonating a leader based on dialogue flow, characterized in that: The system comprises: The acquisition module is used to obtain multiple SMS data; A merging module, used to merge the multiple SMS data into SMS text, and calculate the weight matrix corresponding to the SMS text; The recognition module is used to obtain a text message recognition result according to the text message and a weight matrix corresponding to the text message through a Bi-LSTM model.

9. An electronic device, characterized in that: It includes a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor implements a method for identifying SMS fraud impersonating a leader based on dialogue flow as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements a method for identifying SMS fraud impersonating a leader based on dialogue flow as described in any one of claims 1 to 7.

Images