SMS interception system and method based on AI technology

Abstract

The present invention relates to the technical field of SMS interception, and in particular to an SMS interception system and method based on AI technology. According to the number of word segments corresponding to description texts of different actively intercepted SMS messages, and the number of word segments corresponding to each industry category in all actively intercepted SMS messages, the present invention obtains the SMS interception degree of each user for each industry category, and screens out the preferred industry category for interception of each user; according to the distribution difference of corresponding link intercepted SMS messages between each user and other users, and the similarity of the preferred industry category for interception, the interception feature similarity of actively intercepted SMS messages between each user and other users is obtained, and the interception reference user of each user is screened out; and then the updated blacklist number library and updated interception word segmentation rule library of each user are obtained; and the user's new SMS is intercepted and judged. The present invention improves the accuracy of SMS interception by obtaining the accurate interception word segmentation rule library and blacklist number library of each user.

Description

Technical Field

[0001] The present invention relates to the technical field of SMS interception, and in particular to an SMS interception system and method based on AI technology. Background Art

[0002] SMS is a convenient way of communication and is widely used in various fields. However, due to the continuous proliferation of spam and harassing text messages, it seriously interferes with the normal life and work of users and may even cause economic losses and personal information leakage. Therefore, it is very necessary to intercept spam and harassing text messages.

[0003] In the prior art, text messages are intercepted by judging whether known harassing numbers and keyword contents are contained through methods such as blacklist filtering and keyword filtering. However, due to the subjective interception needs of users, harassing numbers are constantly changing, and with the development and change of network language, new harassing words and expressions will continue to appear. Blacklist filtering cannot intercept text messages sent by new numbers that are not included in the blacklist, and keyword filtering cannot effectively intercept text messages with variant words or no obvious keywords. It fails to learn and adapt to the changing harassing text message patterns in a timely manner, and the effect of text message interception is poor. Summary of the invention

[0004] In order to solve the technical problems that blacklist filtering cannot intercept text messages sent from new numbers that are not included in the blacklist, keyword filtering cannot effectively intercept text messages with variant words or without obvious keywords, fails to learn and adapt to the changing harassing text message patterns in time, and the text message interception effect is poor, the purpose of the present invention is to provide a text message interception system and method based on AI technology, and the technical solutions adopted are as follows: The present invention proposes a text message interception method based on AI technology, the method comprising: Get the word array and phone number of each actively intercepted SMS message of each user; For any actively intercepted SMS, obtain the description text of each segmented word in the segmented word array, and obtain the industry category of each segmented word in the actively intercepted SMS based on the similarity between the description text of each segmented word in the segmented word array and the small category array of each different preset industry category; According to the number of word segments in the description text corresponding to different actively intercepted text messages, and the number of word segments corresponding to each industry category in all actively intercepted text messages, the degree of text message interception of each user for each industry category is obtained; according to the degree of text message interception of each user for each industry category, the preferred industry category for interception of each user is screened out; Obtain link interception text messages from all actively intercepted text messages, and obtain the interception feature similarity of actively intercepted text messages between each user and other users based on the distribution difference of corresponding link interception text messages between each user and other users, and the similarity of interception preferred industry categories; screen out interception reference users for each user based on the interception feature similarity; obtain an updated blacklist number library and an updated interception word segmentation rule library for each user based on all corresponding telephone numbers of each user, and the segmentation corresponding to the interception preferred industry categories between each user and the interception reference user; Based on the number of link intercepted SMS messages for each user, the blacklist number library is updated, and the interception analysis rule library is updated to intercept new SMS messages.

[0005] Furthermore, the method for obtaining the industry category of each segmentation word includes: For any actively intercepted SMS, obtain the description word array of each description text, and obtain the correlation coefficient between the description word array corresponding to each word in the word array and the small category array of each preset industry category as the industry category score value of each word; The industry category with the largest score among the industry categories of each segmented word in the actively intercepted SMS is selected, and the corresponding industry category is used as the industry category of each segmented word.

[0006] Furthermore, the method for obtaining the SMS interception degree includes: For each user, the cumulative value of the number of segmented words corresponding to the description text of all actively intercepted SMS messages is obtained as the number of segmented word descriptions; The ratio of the number of segmented words and the number of segmented word descriptions for each industry category in all actively intercepted text messages is calculated to obtain the degree of text message interception for each industry category for each user.

[0007] Furthermore, the method for obtaining the interception preferred industry category includes: A preset number of industry categories with the highest SMS interception levels for each user are selected, and the corresponding industry categories are used as preferred industry categories for interception.

[0008] Furthermore, the method for obtaining the interception feature similarity includes: Based on the URL regular expression, obtain the link interception SMS in all the actively intercepted SMS; calculate the ratio of the number of link interception SMS corresponding to each user to the number of all actively intercepted SMS as the link proportion; Obtain the difference in link proportion between each user and each other user, and perform negative correlation mapping as the first similarity coefficient; obtain the correlation coefficient of the array composed of intercepted preferred industry categories between each user and each other user as the second similarity coefficient; The first similarity coefficient and the second similarity coefficient are merged to obtain the interception feature similarity for actively intercepting text messages with each other user.

[0009] Furthermore, the method for obtaining the interception reference user includes: If the similarity of interception characteristics of each user actively intercepting text messages with other users is greater than a preset similarity threshold, the corresponding other users will be used as interception reference users for each user.

[0010] Furthermore, the method for acquiring the updated blacklist number library and the updated interception word segmentation rule library of each user includes: Obtain the preset interception word segmentation rule library and blacklist number library for each user; Add the segmentations corresponding to all interception preferred industry categories between each user and the interception reference user to the preset interception segmentation rule library to form an updated interception segmentation rule library; All telephone numbers corresponding to each user are added to the preset blacklist number library to form an updated blacklist number library.

[0011] Furthermore, the method of intercepting new SMS based on the number of link intercepted SMS of each user, updating the blacklist number library and updating the interception analysis rule library includes: If the link ratio of each user is greater than the preset ratio threshold, and the new SMS contains a link, the new SMS is blocked; If the user's link ratio is not greater than the preset ratio threshold, or the new SMS does not contain a link, obtain the phone number corresponding to the new SMS and the similarity between different phone numbers in the updated blacklist number library. If the similarity is greater than the preset similarity, determine to intercept the new SMS; If the similarity is not greater than the preset similarity threshold, obtain the word segmentation array of each new SMS, count the number of word segments in the word segmentation array corresponding to the updated interception word segmentation rule library as the number of matches; calculate the ratio of the number of matches to the number of all word segments as the interception possibility. If the interception possibility is greater than the preset interception threshold, determine to intercept the new SMS.

[0012] Furthermore, the Jaccard correlation coefficient was used to obtain the corresponding correlation coefficient.

[0013] The present invention also proposes an AI-based SMS interception system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any one of the steps of the AI-based SMS interception method.

[0014] The present invention has the following beneficial effects: The present invention obtains the description text of each word in the word segmentation array for any actively intercepted text message, and obtains the industry category of each word in the actively intercepted text message according to the similarity between the description text of each word in the word segmentation array and the small category array of each different preset industry category, thereby realizing the refined classification of the text message content, which is helpful to more accurately understand the intention of the text message content; according to the number of word segments corresponding to the description texts of different actively intercepted text messages, and the number of word segments corresponding to each industry category in all actively intercepted text messages, the degree of text message interception of each user for each industry category is obtained, reflecting the user's interception behavior for text messages of each industry category, and identifying the user's preference and the industry category of key concern; according to the degree of text message interception of each user for each industry category, the preferred industry category for interception of each user is screened out, and the type of text message that the user is most concerned about is clarified; taking into account the active interception of different users There may be a consistent situation in the preference of SMS, obtain the link interception SMS in all active interception SMS, obtain the interception feature similarity of active interception SMS between each user and other users according to the distribution difference of corresponding link interception SMS between each user and other users and the similarity of interception preferred industry categories; according to the interception feature similarity, screen out the interception reference users of each user, find the user group with similar interception behavior, and provide users with more personalized interception suggestions and optimization strategies; according to all telephone numbers corresponding to each user, and the segmentation corresponding to the interception preferred industry category between each user and the interception reference user, obtain the updated blacklist number library and updated interception segmentation rule library of each user, more accurately understand the personalized needs of users, continuously update and optimize the interception information, and realize a more intelligent interception function; interception judgment of new SMS. The present invention obtains the accurate interception segmentation rule library and blacklist number library of each user by considering that the user has subjective interception needs, thereby improving the accuracy of SMS interception. BRIEF DESCRIPTION OF THE DRAWINGS

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

[0016] Figure 1 A flowchart of a method for intercepting short messages based on AI technology provided by one embodiment of the present invention; Figure 2 A flow chart of a method for obtaining interception feature similarity provided by an embodiment of the present invention; Figure 3 A flow chart of a method for obtaining a new SMS interception judgment provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation, structure, features and effects of a text message interception system and method based on AI technology proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0018] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0019] The following is a detailed description of a specific scheme of a text message interception system and method based on AI technology provided by the present invention in conjunction with the accompanying drawings.

[0020] See also Figure 1 , which shows a flowchart of a text message interception method based on AI technology provided by an embodiment of the present invention, specifically comprising: Step S1: Obtain the word segmentation array and phone number of each actively intercepted SMS of each user.

[0021] In the embodiments of the present invention, taking into account that users usually have subjective needs, there is a situation where text messages that are not in the preset interception word segmentation rule library are actively intercepted, resulting in different actively intercepted text messages for each user, and different keywords in the corresponding text messages. Therefore, in order to improve the accuracy of intelligent interception of text messages, the present invention further analyzes the user's actively intercepted text messages, adjusts the information in the interception word segmentation rule library and the blacklist number library, so that subsequent interception can be carried out in a timely manner, reducing the number of manual interceptions by users; first, the actively intercepted text messages are segmented based on the Jieba word segmentation tool to obtain the word segmentation array and telephone number of each actively intercepted text message of each user.

[0022] Jieba word segmentation can accurately identify word boundaries in Chinese texts, especially for new words. Jieba word segmentation is a technical means well known to those skilled in the art, and will not be described in detail here.

[0023] It should be noted that, in one embodiment of the present invention, the time range for actively intercepting text messages is one week. In other embodiments of the present invention, the time range for actively intercepting text messages can be set according to specific circumstances, which is not limited or elaborated here.

[0024] Step S2: For any actively intercepted SMS, obtain the description text of each segmented word in the segmented word array, and obtain the industry category of each segmented word in the actively intercepted SMS based on the similarity between the description text of each segmented word in the segmented word array and the small category array of each different preset industry category.

[0025] Since the content of SMS often contains words related to a specific industry, by segmenting and obtaining the description text of each segmentation, the specific meaning of the segmentation in the context is provided, and the subject and intention of the SMS can be more accurately identified. For any actively intercepted SMS, the description text of each segmentation in the segmentation array is obtained. It should be noted that in one embodiment of the present invention, the description text of the segmentation on Baidu Encyclopedia is obtained by a web crawler. The specific web crawler is a technical means well known to those skilled in the art and will not be described in detail here.

[0026] Each word segment in the SMS content may be associated with multiple industry categories. By further analyzing the word segmentation description array, richer contextual information can be provided to help the system identify the industry attributes of the word segmentation more accurately. Since different industry categories have different characteristics and key words, by comparing the similarity between the word segmentation description array and each small category array, the SMS content can be refined. Therefore, based on the similarity between the description text of each word segment in the word segmentation array and the small category array of each different preset industry category, the industry category of each word segment in the actively intercepted SMS is obtained.

[0027] Preferably, in one embodiment of the present invention, the method for obtaining the industry category of each segmented word includes: For any actively intercepted SMS, obtain the description word array of each description text, obtain the correlation coefficient between the description word array corresponding to each word in the word array and the small category array of each preset industry category, as the industry category score value of each word; The industry category with the largest score among the industry categories of each segmented word in the actively intercepted SMS is selected, and the corresponding industry category is used as the industry category of each segmented word.

[0028] In one embodiment of the present invention, the formula for the industry category score is expressed as: ; in, Indicates The first word in the word segmentation array of the active interception text message The first participle Industry category score value; Indicates The first word in the word segmentation array of the active interception text message The description word array of word segments; Indicates An array of small categories of preset industry categories; represents the Jaccard correlation coefficient.

[0029] In the formula of the industry category score value, the larger the correlation coefficient is, the more similar the content of the description word array of each word is to the small category array of each preset industry category, and the more likely the content of the word corresponds to the industry category. It should be noted that, in one embodiment of the invention, the preset industry categories can be obtained in advance by the implementers by referring to relevant professional literature, that is, the standard industry categories divided by the country according to the principle of homogeneity of economic activities, which will not be elaborated here.

[0030] It should be noted that, in one embodiment of the present invention, after obtaining the description word segmentation array of the description text through Jieba word segmentation, the Jaccard correlation coefficient is used to calculate the similarity between the two arrays to obtain the correlation coefficient between the description word segmentation array of each word segmentation and the small category array of each preset industry category. The specific Jieba word segmentation and Jaccard correlation coefficient are technical means well known to those skilled in the art and will not be elaborated here.

[0031] Step S3: According to the number of word segments corresponding to the description texts of different actively intercepted text messages, and the number of word segments corresponding to each same industry category in all actively intercepted text messages, the degree of text message interception of each user for each industry category is obtained; according to the degree of text message interception of each user for each industry category, the preferred industry category for interception of each user is screened out.

[0032] The number of word segments in the description text corresponding to the actively intercepted SMS can reflect the user's attention to the content that shows the characteristics of the SMS information. The more word segments in the corresponding description text, the more complex and informative the SMS content is. By analyzing the number of word segments in each industry category, the proportion of SMS in each industry category in the overall interception is reflected. The larger the proportion, the greater the degree of SMS interception by users for the industry category. Therefore, based on the number of word segments in the description text corresponding to different actively intercepted SMS, and the number of word segments corresponding to each industry category in all actively intercepted SMS, the degree of SMS interception of each user for each industry category is obtained.

[0033] Preferably, in one embodiment of the present invention, the method for obtaining the SMS interception degree includes: For each user, the cumulative value of the number of segmented words corresponding to the description text of all actively intercepted SMS messages is obtained as the number of segmented word descriptions; The ratio of the number of segmented words and the number of segmented word descriptions for each industry category in all actively intercepted text messages is calculated to obtain the degree of text message interception for each industry category for each user.

[0034] In one embodiment of the present invention, the formula for the SMS interception degree is expressed as: ; in, Indicates User to The degree of SMS interception by industry category; Indicates The user corresponds to the first The number of word segments in each industry category; Indicates The number of word segments corresponding to the description text of the actively intercepted SMS; Indicates the number of actively intercepted SMS messages.

[0035] In the formula for SMS interception degree, Indicates the cumulative value of the number of segmented words corresponding to the description text of all actively intercepted SMS messages, that is, the number of segmented word descriptions. Indicates the number of all actively intercepted text messages. The ratio of the number of segmentation words and the number of segmentation descriptions of each industry category. The larger the number of segmentation descriptions, the higher the corresponding The more word segmentations there are for an industry category, the more interceptions there are for an industry category, and the greater the degree of SMS interception.

[0036] Since users are more inclined to block text messages that are irrelevant, uninteresting or frequently disturbing, the degree of text message blocking can reflect each user's interception behavior of text messages in each industry category, which helps to understand the user's real needs and preferences. The greater the degree of text message blocking, the more frequently the user intercepts text messages in the corresponding industry category, and the more targeted analysis of the industry category is needed; based on each user's degree of text message blocking in each industry category, the preferred industry categories for interception are screened out for each user.

[0037] Preferably, in one embodiment of the present invention, the method for intercepting and obtaining the preferred industry category includes: A preset number of industry categories with the highest SMS interception levels for each user are selected, and the corresponding industry categories are used as preferred industry categories for interception.

[0038] It should be noted that, in one embodiment of the present invention, the preset number is 3, and the 3 highest SMS interception levels are selected, and the corresponding industry categories are the preferred industry categories for interception; in other embodiments of the present invention, the size of the preset number can be set according to the specific circumstances, and is not limited or elaborated here.

[0039] Step S4: Obtain link interception text messages from all actively intercepted text messages, and obtain the interception feature similarity of actively intercepted text messages between each user and other users based on the distribution difference of corresponding link interception text messages between each user and other users, and the similarity of interception preferred industry categories; screen out interception reference users for each user based on the interception feature similarity; obtain each user's updated blacklist number library and updated interception word segmentation rule library based on all telephone numbers corresponding to each user, and the segmentation corresponding to the interception preferred industry categories between each user and the interception reference user.

[0040] Link interception SMS reflects the active interception of SMS containing malicious links or fraudulent content. This is done by analyzing the user's interception behavior patterns, that is, the behavior of intercepting link SMS and interception preferred industry categories. The more consistent the interception behavior and the corresponding industry categories are, the greater the similarity of the interception characteristics. Therefore, the link interception SMS among all the active interception SMS are obtained, and based on the distribution differences of the corresponding link interception SMS between each user and other users, and the similarity of the interception preferred industry categories, the similarity of the interception characteristics of the active interception SMS between each user and other users is obtained.

[0041] Preferably, in one embodiment of the present invention, the method for obtaining interception feature similarity can be found in Figure 2 , which shows a flow chart of a method for obtaining interception feature similarity, including: Step S201: obtaining link interception text messages among all actively intercepted text messages based on URL regular expressions; calculating the ratio of the number of link interception text messages corresponding to each user to the number of all actively intercepted text messages as the link proportion.

[0042] Since the link in the text message is a clickable web page used to guide the recipient to visit a specific web page, it is often associated with risks such as fraud and malware transmission. Therefore, it is necessary to analyze the link text messages. Usually the link appears in the form of a URL, which is an existing standard format. The regular expression of the URL can match most or specific types of URL formats to obtain link interception text messages in all active interception text messages; the link proportion reflects the frequency of users receiving interception text messages containing links. The larger the link proportion, the greater the possibility that the link is a harassing text message.

[0043] It should be noted that the specific URL regular expression is a technical means well known to those skilled in the art and will not be described in detail here.

[0044] Step S202: Obtain the difference in link proportion between each user and each other user, and perform negative correlation mapping as the first similarity coefficient; obtain the correlation coefficient of the array composed of intercepted preferred industry categories between each user and each other user as the second similarity coefficient.

[0045] The difference in link proportion between each user and every other user can reflect whether the users have similar concerns about link interception SMS. The larger the difference, the more inconsistent the behavior of link SMS interception. Therefore, it is necessary to negatively correlate the differences to obtain the first similarity coefficient. The larger the performance difference, the smaller the similarity of users' SMS interception behavior patterns. Since the industry category of SMS is one aspect of analyzing user interception behavior patterns, the correlation coefficient of the array composed of the interception preferred industry categories between each user and every other user is analyzed to reflect whether the industry categories of interception preferences between users are similar. The larger the correlation coefficient, the more consistent the industry categories of interception preferences, and the larger the second similarity coefficient.

[0046] Step S203: The first similarity coefficient and the second similarity coefficient are merged to obtain the interception feature similarity of actively intercepting short messages between each user and each other user.

[0047] The larger the first similarity coefficient is, the smaller the difference in the proportion of links is, the more consistent the interception behavior is for link text messages, and the greater the interception similarity is; the larger the second similarity coefficient is, the greater the correlation coefficient of the array composed of the preferred industry categories for interception between users is, and the more similar the interception industry category characteristics are. Therefore, the larger the first similarity coefficient is, the larger the second similarity coefficient is, and the greater the similarity of the interception characteristics is.

[0048] It should be noted that, in some embodiments of the present invention, the first similarity coefficient and the second similarity coefficient may be fused by addition or multiplication. The specific means are well known to those skilled in the art and will not be elaborated herein.

[0049] In one embodiment of the present invention, the formula for intercept feature similarity is expressed as: ; in, Indicates Users and The similarity of interception characteristics of active SMS interception between users; Indicates The number of actively intercepted SMS messages containing link information by each user; Indicates The number of actively blocked SMS messages per user; Indicates The number of actively intercepted SMS messages containing link information by each user; Indicates The number of actively blocked SMS messages per user; Indicates An array of interception preferred industry categories for each user; Indicates An array of interception preferred industry categories for each user; Indicates Users and The correlation coefficient of the array composed of the interception of the preferred industry categories between users, that is, the second similarity coefficient; represents an exponential function with a natural constant as base; Indicates taking the absolute value.

[0050] In the formula for intercept feature similarity, It means that the exponential function with the natural constant as the base is used to convert Perform negative correlation mapping, i.e. the first similarity coefficient; Indicates the calculation The number of active intercepted SMS messages containing link information by users and The ratio of all actively intercepted SMS messages of a user, i.e. The link share of each user; Indicates the calculation The number of active intercepted SMS messages containing link information by users and The ratio of all actively intercepted SMS messages of a user, i.e. The link share of each user, It means calculating the difference in the proportion of links between each user and every other user. The larger the difference, the more inconsistent the interception preference of SMS containing links, and the worse the similarity of interception characteristics. Users and The larger the correlation coefficient of the array composed of the intercepted preferred industry categories between users, the more consistent the intercepted preferred industry categories between users, the more information in the same industry is intercepted, and the greater the similarity of the intercepted features.

[0051] It should be noted that, in one embodiment of the present invention, by The function will A negative correlation mapping is performed to form a negative correlation relationship in which the greater the difference, the smaller the similarity of the intercept characteristics; in other embodiments of the present invention, The negative correlation is realized by finding the reciprocal. If the reciprocal method is adopted, a threshold value, such as 0.01, needs to be added to the denominator to avoid the denominator of the formula being 0, which makes the formula meaningless. The specific means are well known to those skilled in the art and will not be elaborated here.

[0052] The similarity of interception characteristics reflects that different users may make similar interception decisions when facing the same or similar types of SMS. By finding users with similar interception characteristics as interception references, more personalized interception suggestions and optimization strategies can be provided to users, which can more accurately reflect the actual needs of users and improve the satisfaction and effect of interception services. Therefore, based on the similarity of interception characteristics, each user's interception reference user is screened out.

[0053] Preferably, in one embodiment of the present invention, the method for obtaining the interception reference user includes: If the similarity of interception characteristics of each user actively intercepting text messages with other users is greater than a preset similarity threshold, the corresponding other users will be used as interception reference users for each user.

[0054] It should be noted that, in one embodiment of the present invention, the size of the preset similarity threshold is 0.7. In other embodiments of the present invention, the size of the preset similarity threshold can be set according to specific circumstances, which is not limited or elaborated here.

[0055] Since users have subjective needs, different users have different sensitivities and preferences for harassing calls and text messages. As time goes by and user behavior changes, by analyzing the interception behavior of individual users and their interception reference users, we can more accurately understand the personalized needs of users, continuously update and optimize the interception information, and achieve a more intelligent interception function; therefore, based on all the phone numbers corresponding to each user, and the segmentation corresponding to the interception preferred industry category between each user and the interception reference user, we can obtain each user's updated blacklist number library and updated interception segmentation rule library.

[0056] Preferably, in one embodiment of the present invention, the method for acquiring the updated blacklist number library and the updated interception word segmentation rule library of each user includes: Obtain the preset interception word segmentation rule library and blacklist number library for each user; Add the segmentations corresponding to all interception preferred industry categories between each user and the interception reference user to the preset interception segmentation rule library to form an updated interception segmentation rule library; All telephone numbers corresponding to each user are added to the preset blacklist number library to form an updated blacklist number library.

[0057] It should be noted that, in the embodiments of the present invention, the preset interception word segmentation rule library and the blacklist number library can be pre-acquired by the implementation personnel based on the existing relevant professional materials and data, which will not be elaborated here.

[0058] Step S5: Based on the number of link intercepted SMS messages of each user, the blacklist number library is updated, and the interception analysis rule library is updated to make interception judgments on new SMS messages.

[0059] Based on the number of link-intercepted SMS messages for each user, personalized interception can more accurately identify and intercept SMS messages that users do not need, reducing the number of false interceptions and missed interceptions; updating the blacklist number library and interception analysis rule library can ensure that the system can respond to new fraud methods and spam SMS messages in a timely manner, meet personalized interception needs, and improve the system's protection capabilities and response speed. Based on the number of link-intercepted SMS messages for each user, update the blacklist number library, and update the interception analysis rule library to intercept new SMS messages.

[0060] Preferably, in one embodiment of the present invention, based on the number of link intercepted SMS messages for each user, the blacklist number library is updated, and the interception analysis rule library is updated to determine the interception of new SMS messages. Figure 3 , which shows a flow chart of a method for obtaining a new SMS interception judgment, including: Step S301: If the link ratio of each user is greater than a preset ratio threshold, and the new SMS has a link, it is determined to intercept the new SMS.

[0061] It should be noted that, in one embodiment of the present invention, the size of the preset proportion threshold is 2 / 3. In other embodiments of the present invention, the size of the preset proportion threshold can be set according to specific circumstances, which is not limited or elaborated here.

[0062] Step S302: If the user's link ratio is not greater than the preset ratio threshold, or the new text message does not have a link, obtain the phone number corresponding to the new text message and the similarity between different phone numbers in the updated blacklist number library. If the similarity is greater than the preset similarity, determine to intercept the new text message.

[0063] It should be noted that, considering that the blacklist number library contains known malicious numbers, as long as there is a certain degree of similarity, it may be a potential spam or fraudulent text message. In one embodiment of the present invention, the similarity can be obtained according to the cosine similarity, and the preset similarity is 0; in other embodiments of the present invention, the preset similarity can be analyzed according to the specific situation, and it is not limited or elaborated here. The specific cosine similarity is a technical means well known to those skilled in the art, and it is not elaborated here.

[0064] Step S303: If the similarity is not greater than the preset similarity, obtain the word segmentation array of each new SMS, count the number of word segments in the word segmentation array corresponding to the updated interception word segmentation rule library as the matching number; calculate the ratio of the matching number to the number of all word segments as the interception possibility. If the interception possibility is greater than the preset interception threshold, determine to intercept the new SMS.

[0065] The number of matches is reflected in the number of segmented words in the updated interception segmentation rule library. The more segmented words there are, the more information the new SMS contains that needs to be intercepted, and the greater the possibility of interception.

[0066] It should be noted that, in one embodiment of the present invention, the size of the preset interception threshold is 0.5. In other embodiments of the present invention, the size of the preset interception threshold can be set according to specific circumstances, which is not limited or elaborated here.

[0067] In summary, the present invention obtains the SMS interception degree of each user for each industry category according to the number of segmentations of the description text corresponding to different active intercepted SMS, and the number of segmentations corresponding to each industry category in all active intercepted SMS, and screens out the preferred industry category for interception of each user; obtains the interception feature similarity of active intercepted SMS between each user and other users according to the distribution difference of corresponding linked intercepted SMS between each user and other users, and the similarity of intercepted preferred industry category, and screens out the interception reference user of each user; and then obtains the updated blacklist number library and updated interception segmentation rule library of each user; and intercepts and judges the new SMS. The present invention improves the accuracy of SMS interception by obtaining the accurate interception segmentation rule library and blacklist number library of each user.

[0068] The present invention also proposes an AI-based SMS interception system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any one of the steps of an AI-based SMS interception method.

[0069] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0070] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A SMS interception method based on AI technology, It is characterized in that The method comprises: Get the word array and phone number of each actively intercepted SMS message of each user; For any actively intercepted SMS, obtain the description text of each segmented word in the segmented word array, and obtain the industry category of each segmented word in the actively intercepted SMS based on the similarity between the description text of each segmented word in the segmented word array and the small category array of each different preset industry category; According to the number of word segments in the description text corresponding to different actively intercepted text messages, and the number of word segments corresponding to each industry category in all actively intercepted text messages, the degree of text message interception of each user for each industry category is obtained; according to the degree of text message interception of each user for each industry category, the preferred industry category for interception of each user is screened out; Obtain link interception text messages from all actively intercepted text messages, and obtain the interception feature similarity of actively intercepted text messages between each user and other users based on the distribution difference of corresponding link interception text messages between each user and other users, and the similarity of interception preferred industry categories; screen out interception reference users for each user based on the interception feature similarity; obtain an updated blacklist number library and an updated interception word segmentation rule library for each user based on all corresponding telephone numbers of each user, and the segmentation corresponding to the interception preferred industry categories between each user and the interception reference user; Based on the number of link intercepted SMS messages for each user, the blacklist number library is updated, and the interception analysis rule library is updated to intercept new SMS messages.

2. According to claim 1, a text message interception method based on AI technology, It is characterized in that The method for obtaining the industry category of each segmentation word includes: For any actively intercepted SMS, obtain the description word array of each description text, and obtain the correlation coefficient between the description word array corresponding to each word in the word array and the small category array of each preset industry category as the industry category score value of each word; The industry category with the largest score among the industry categories of each segmented word in the actively intercepted SMS is selected, and the corresponding industry category is used as the industry category of each segmented word.

3. According to claim 1, a text message interception method based on AI technology, It is characterized in that The method for obtaining the SMS interception degree includes: For each user, the cumulative value of the number of segmented words corresponding to the description text of all actively intercepted SMS messages is obtained as the number of segmented word descriptions; The ratio of the number of segmented words and the number of segmented word descriptions for each industry category in all actively intercepted text messages is calculated to obtain the degree of text message interception for each industry category for each user.

4. According to the method for intercepting short messages based on AI technology according to claim 1, It is characterized in that The method for obtaining the interception preferred industry category includes: A preset number of industry categories with the highest SMS interception levels for each user are selected, and the corresponding industry categories are used as preferred industry categories for interception.

5. According to claim 1, a text message interception method based on AI technology, It is characterized in that The method for obtaining the interception feature similarity includes: Based on the URL regular expression, obtain the link interception SMS in all the actively intercepted SMS; calculate the ratio of the number of link interception SMS corresponding to each user to the number of all actively intercepted SMS as the link proportion; Obtain the difference in link proportion between each user and each other user, and perform negative correlation mapping as the first similarity coefficient; obtain the correlation coefficient of the array composed of intercepted preferred industry categories between each user and each other user as the second similarity coefficient; The first similarity coefficient and the second similarity coefficient are merged to obtain the interception feature similarity for actively intercepting text messages with each other user.

6. According to claim 1, a text message interception method based on AI technology, It is characterized in that The method for obtaining the interception reference user includes: If the similarity of interception characteristics of each user actively intercepting text messages with other users is greater than a preset similarity threshold, the corresponding other users will be used as interception reference users for each user.

7. According to claim 1, a method for intercepting short messages based on AI technology, It is characterized in that The method for obtaining the updated blacklist number library and the updated interception word segmentation rule library of each user includes: Obtain the preset interception word segmentation rule library and blacklist number library for each user; Add the segmentations corresponding to all interception preferred industry categories between each user and the interception reference user to the preset interception segmentation rule library to form an updated interception segmentation rule library; All telephone numbers corresponding to each user are added to the preset blacklist number library to form an updated blacklist number library.

8. According to claim 5, a method for intercepting short messages based on AI technology, It is characterized in that The method of intercepting new SMS based on the number of link intercepted SMS of each user, updating the blacklist number library, and updating the interception analysis rule library includes: If the link ratio of each user is greater than the preset ratio threshold, and the new SMS contains a link, the new SMS is blocked; If the user's link ratio is not greater than the preset ratio threshold, or the new SMS does not contain a link, obtain the phone number corresponding to the new SMS and the similarity between different phone numbers in the updated blacklist number library. If the similarity is greater than the preset similarity, determine to intercept the new SMS; If the similarity is not greater than the preset similarity threshold, obtain the word segmentation array of each new SMS, count the number of word segments in the word segmentation array corresponding to the updated interception word segmentation rule library as the number of matches; calculate the ratio of the number of matches to the number of all word segments as the interception possibility. If the interception possibility is greater than the preset interception threshold, determine to intercept the new SMS.

9. According to the method for intercepting short messages based on AI technology according to claim 2 or 5, It is characterized in that The Jaccard correlation coefficient was used to obtain the corresponding correlation coefficient.

10. A text message interception system based on AI technology, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the computer program, the steps of the SMS interception method based on AI technology as described in any one of claims 1 to 9 are implemented.

Images