Fraud-related group identification method and device, storage medium and electronic equipment

By screening web pages and application traffic from network traffic, determining the target fraud platform, extracting user behavior characteristics and calculating risk levels, the problem of difficult to identify fraud-related groups in the existing technology is solved, and the accurate identification and tracking of fraud-related groups is achieved , improves identification efficiency and accuracy.

CN119004340APending Publication Date: 2024-11-22BEIJING FULE TECH CO LTD

Patent Information

Application Number
CN202411462685.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-19
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

It is difficult for the prior art to accurately identify fraud-related groups, especially online fraud traffic with diversified patterns, complex processes, professional tools and strong concealment, resulting in the inability to effectively identify the behavioral characteristics of fraud-related groups.

Method used

By filtering web page traffic and application traffic from network traffic, determining the target fraud platform, extracting user behavior characteristics, matching target tags, calculating risk levels, and tracking the traffic trends of suspected fraud users based on abnormal network fingerprints, realizing the fraud-related groups Accurate identification.

Benefits of technology

Accurate identification and tracking of fraud-related groups has been achieved, the accuracy and efficiency of identification have been improved, high-risk behaviors can be quickly identified, and fraud cases can be prevented.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119004340A_ABST
    Figure CN119004340A_ABST
Patent Text Reader

Abstract

The invention relates to a fraud-related group identification method and device, a storage medium and electronic equipment, and relates to the technical field of network security, and the method comprises the following steps: screening webpage traffic and application traffic from network traffic, determining at least one to-be-evaluated webpage corresponding to the webpage traffic, and determining at least one to-be-evaluated application corresponding to the application traffic; determining a target fraud platform from each to-be-evaluated webpage and each to-be-evaluated application, and extracting user behavior characteristics of a target user in the target fraud platform; based on the at least one user behavior feature of the same target user, determining a target tag of the corresponding target user from a preset user tag library; calculating a first risk level of the corresponding target user as a fraud-related group based on the target label corresponding to the same target user; and when the first risk level is greater than a preset level threshold, determining the corresponding target user as a suspected fraud-related user. The method and the device have the effect of accurately identifying the fraud-related group.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of network security technology, and specifically to a method, device, storage medium and electronic device for identifying a group involved in fraud. Background Art

[0002] Fraudulent groups refer to individuals or groups that participate in or are suspected of participating in fraudulent activities. They are a serious threat to public security and financial security, causing huge economic losses and psychological harm to society and individuals. In daily life, the most common fraudulent activity is Internet fraud, which is a criminal act of using the Internet and other communication tools to fabricate or distort facts to defraud others of their property. In order to effectively prevent and combat fraudulent activities, it is necessary to conduct in-depth analysis and research on the characteristics and patterns of fraudulent groups, so as to improve the ability to identify fraudulent groups to a certain extent.

[0003] At present, the usual method for identifying fraud-related groups is to extract the user's current behavior pattern from the network traffic, and through the similarity between the user's current behavior pattern and the historical behavior pattern, to determine if the user's behavior is abnormal, and then identify whether it is a fraud-related group. However, for network fraud traffic with characteristics such as diversified patterns, complicated processes, specialized tools, and strong concealment, it is difficult to extract the behavioral characteristics of the fraud-related group from it, resulting in the inability to accurately identify the fraud-related group. Summary of the invention

[0004] In order to accurately identify groups involved in fraud, the present application provides a method, device, storage medium and electronic device for identifying groups involved in fraud.

[0005] In a first aspect of the present application, a method for identifying a fraudulent group is provided, which specifically includes: Get network traffic; Filtering web page traffic and application traffic from the network traffic, and determining at least one web page to be evaluated corresponding to the web page traffic, and determining at least one application to be evaluated corresponding to the application traffic; Determine a target fraud platform from each of the web pages to be evaluated and each of the applications to be evaluated, and extract user behavior characteristics of a target user in the target fraud platform, wherein the target user is a user who accesses and operates the target fraud platform; Based on at least one user behavior feature of the same target user, determine a target tag of the corresponding target user from a preset user tag library, wherein the user tag library includes at least one tag possessed by a fraud-related group; Based on the target tag corresponding to the same target user, calculating the first risk level of the corresponding target user as a fraud-related group, wherein the higher the first risk level, the higher the possibility that the corresponding target user is a fraud-related group; When the first risk level is greater than a preset level threshold, the corresponding target user is determined as a suspected fraud user, and the corresponding first abnormal network fingerprint is determined based on the user traffic of the suspected fraud user. Based on the first abnormal network fingerprint, the traffic trend of the corresponding suspected fraud user is tracked.

[0006] By adopting the above technical solution, network traffic is obtained, web traffic and application traffic are screened out from the network traffic, and then the web page to be evaluated from which the web traffic comes is determined, that is, the web page that needs to be evaluated whether it is a fraudulent web page, and the application to be evaluated from which the application traffic comes is determined, that is, the application that needs to be evaluated whether it is a fraudulent application. Further, the target fraud platform is screened out from each web page to be evaluated and each application to be evaluated, and then the user behavior characteristics of the target user in the target fraud platform are used to match the corresponding target user with an adapted target tag, and according to the target tag, the first risk level of the target user as a fraudulent group is determined, so as to accurately evaluate the possibility that the target user is a fraudulent group. If the first risk level is greater than the level threshold, it means that the corresponding target user is likely to be a fraudulent group, then it is determined as a suspected fraudulent user, and the accurate identification of the fraudulent group is achieved. Finally, according to the first abnormal network fingerprint, the traffic trend of the suspected fraudulent user is tracked, so that the subsequent high-risk behavior of the fraudulent group can be quickly and accurately identified.

[0007] Optionally, determining the target tag of the corresponding target user from a preset user tag library based on at least one user behavior feature of the same target user specifically includes: Calculate the abnormal score value of the user behavior characteristics of the same target user through a preset abnormal score value calculation formula; Based on each abnormal score value corresponding to the same target user, obtain an abnormal behavior feature vector of the corresponding target user on the target fraud platform; According to each abnormal score value in the abnormal behavior feature vector, a corresponding adaptation tag is matched from a preset user tag library, and each adaptation tag is determined as a target tag of the corresponding target user.

[0008] By adopting the above technical solution, the abnormal score value of each user behavior feature of the target user is calculated through the abnormal score value calculation formula, thereby characterizing the abnormal degree of each user behavior feature, and finally determining the abnormal behavior feature vector of the target user. Further, according to each abnormal score value in the abnormal behavior feature vector of a single target user, the label of the fraud-related group, that is, the target label, is accurately matched for this target user, so that it is convenient to accurately determine whether the target user is a fraud-related group based on the target label.

[0009] Optionally, the abnormality score value calculation formula is: s=Wf ·( 100)+W t ·( 100) + W i ·( 100)+W p ·( 100)+W a ·( 100); Where W f , W t , W i , W p and W a All represent fixed weights, the sum of all fixed weights is 1, f observed The behavior frequency representing the user behavior characteristics in the target fraud platform, f threshold represents the preset behavior frequency threshold, t observed represents the duration of user behavior characteristics in the target fraud platform, t threshold represents the preset duration threshold, i observed represents the number of interactions of the target user in the target fraud platform, i threshold represents the preset interaction number threshold, p observed represents the payment amount of the target user on the target fraud platform, p threshold Indicates the preset payment amount threshold, a observed Indicates the number of occurrences of the same type of behavior characteristics as the user behavior characteristics, a threshold represents the preset occurrence threshold, and s represents the abnormal score value of the user behavior feature.

[0010] By adopting the above technical solution, the abnormal score value calculation formula performs weighted summation from multiple dimensions such as behavior frequency, behavior duration, number of interactions, payment behavior, etc., so as to more comprehensively and accurately determine the abnormal score value of the user behavior characteristics of the target user, and then more accurately determine the abnormal degree of the target user's behavior on the target fraud platform, so as to facilitate the subsequent matching of corresponding target tags for corresponding target users according to user behavior characteristics.

[0011] Optionally, the step of calculating the first risk level of the corresponding target user as a fraud-related group based on the target tag corresponding to the same target user specifically includes: Determine each abnormal score value of the same target user as the label value of the corresponding target label, and perform weighted summation on the label values ​​to obtain the corresponding target user's fraud risk score; Substitute the fraud risk score into the preset risk level calculation formula to obtain the first risk level of the corresponding target user as a fraud group. The risk level calculation formula is: ; In the formula, R represents the first risk level, S represents the fraud risk score, and P represents the possibility of the target user committing fraud, ranging from 0 to 1. Indicates the number of times the target user has triggered the risk within the latest preset time. It indicates the total number of times the target user has triggered risks in history, and T indicates the duration corresponding to the preset time. and All represent adjustment parameters. Represents the scaling parameter.

[0012] By adopting the above technical solution, the product of S and P reflects the probability of the target user currently committing fraud. The ratio to T is used to measure the frequency of recent high-risk behaviors of target users. Reflects the cumulative effect of the target users’ long-term high-risk behaviors. Taking the logarithm can effectively avoid over-amplifying the impact of historical high-risk behaviors, thereby ultimately making the first risk level of the determined target user more accurate.

[0013] Optionally, determining the corresponding first abnormal network fingerprint according to the user traffic of the suspected fraudulent user specifically includes: Extracting at least one device feature information from the user traffic of the suspected fraudulent user; Determining a feature type of each device feature information, and dividing each device feature information into multiple levels according to the feature type; Aggregate and calculate the characteristic information of each device in each layer to obtain the corresponding intra-layer aggregation result; The aggregation results within each layer are weightedly summed to obtain an inter-layer combination result, and the inter-layer combination result is hashed using a preset hash algorithm to obtain a first abnormal network fingerprint corresponding to the suspected fraud-related user.

[0014] By adopting the above technical solution, the feature information of each device is divided into multiple levels according to the feature type. Furthermore, the feature information of each device is first aggregated and calculated within the corresponding level, and then the aggregation results within different levels are combined between levels to finally obtain the first abnormal network fingerprint, thereby realizing independent calculation and analysis of device features at different levels, reducing the impact of device feature anomalies within a single level on the determination of network fingerprints, and thereby improving the accuracy of the subsequent first abnormal network fingerprint.

[0015] Optionally, the user tag library includes at least one tag possessed by the defrauded group, and the method further includes: When the target tags corresponding to the same target user are all tags possessed by the fraud-stricken group, the second risk level of the corresponding target user as the fraud-stricken group is calculated according to each of the target tags; When the second risk level is greater than a preset level threshold, the corresponding target user is determined as a suspected defrauded user, and a corresponding second abnormal network fingerprint is determined according to the user traffic of the suspected defrauded user; According to the second abnormal network fingerprint, it is determined whether the corresponding suspected defrauded user is accessing a fraud platform. If so, an anti-fraud reminder message is sent to the device of the corresponding suspected defrauded user.

[0016] By adopting the above technical solution, if the target tags corresponding to the same target user are all tags possessed by the fraud group, it means that this target user is not likely to be a fraud group. Then, according to each target tag, the second risk level of the corresponding target user as a fraud group is calculated. If the second risk level is greater than the preset level threshold, it means that the corresponding target user is likely to be a fraud group. Then the corresponding target user is determined as a suspected fraud user. Furthermore, if the corresponding suspected fraud user is tracked to be accessing a fraud platform according to the second abnormal network fingerprint, anti-fraud reminder information is sent to the device of the corresponding suspected fraud user, thereby curbing the occurrence of fraud cases to a certain extent.

[0017] Optionally, filtering web page traffic and application traffic from the network traffic specifically includes: Preprocessing the network traffic to obtain processed network traffic; Extracting application layer data from the processed network traffic through deep packet inspection, and determining a first weight and a second weight of each flow feature information in the application layer data, wherein the first weight is a characteristic weight of the corresponding flow feature information in the webpage, and the second weight is a characteristic weight of the corresponding flow feature information in the application; According to the first weight, weighted sum of the first scores of each piece of traffic feature information is performed to obtain a first score, and according to the second weight, weighted sum of the second scores of each piece of traffic feature information is performed to obtain a second score; If the first score is greater than the second score, the corresponding part of the traffic is determined as web page traffic; if the first score is less than the second score, the corresponding part of the traffic is determined as application traffic.

[0018] By adopting the above technical solution, if the first score is greater than the second score, it means that the traffic is more likely to come from the web page, and then the corresponding part of the traffic is determined to be web page traffic; if the second score is greater than the first score, it means that the traffic is more likely to come from the application, and then the corresponding part of the traffic is determined to be application traffic, thereby realizing the screening of web page traffic and application traffic in network traffic.

[0019] In a second aspect of the present application, a fraud-related group identification device is provided, which specifically includes: Traffic capture module, used to obtain network traffic; A traffic screening module, used to screen web page traffic and application traffic from the network traffic, and determine at least one web page to be evaluated corresponding to the web page traffic, and determine at least one application to be evaluated corresponding to the application traffic; A behavior extraction module, used to determine a target fraud platform from each of the web pages to be evaluated and each of the applications to be evaluated, and to extract user behavior characteristics of a target user in the target fraud platform, wherein the target user is a user who accesses and operates the target fraud platform; A tag matching module, used to determine a target tag of a corresponding target user from a preset user tag library based on at least one user behavior feature of the same target user, wherein the user tag library includes at least one tag possessed by a fraud-related group; A risk determination module, configured to calculate, based on a target tag corresponding to the same target user, a first risk level of the corresponding target user being a fraud-related group, wherein the higher the first risk level, the higher the possibility that the corresponding target user is a fraud-related group; The group identification module is used to determine the corresponding target user as a suspected fraud user when the first risk level is greater than a preset level threshold, determine the corresponding first abnormal network fingerprint based on the user traffic of the suspected fraud user, and track the traffic trend of the corresponding suspected fraud user based on the first abnormal network fingerprint.

[0020] By adopting the above technical solution, the traffic capture module obtains the network traffic, the traffic screening module screens out the web page traffic and application traffic from the network traffic, and determines at least one web page to be evaluated and application to be evaluated, and then the behavior extraction module extracts the user behavior characteristics of the target user in the target fraud platform, and then the label matching module matches the corresponding target label for the target user. Furthermore, the risk determination module determines the first risk level of the corresponding target user as a fraud-related group based on the target label, and finally the group identification module determines the first abnormal network fingerprint corresponding to the suspected fraud-related user when the first risk level is greater than the level threshold, and tracks the traffic trend of the suspected fraud-related user based on the first abnormal network fingerprint.

[0021] In a third aspect of the present application, a computer-readable storage medium is provided, in which a computer program is stored. When the computer program is loaded and executed by a processor, the method steps described in any one of the first aspects are performed.

[0022] In a fourth aspect of the present application, an electronic device is provided, specifically comprising: A processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the processor is used to load and execute the computer program stored in the memory so that the electronic device performs the method as described in any one of the first aspects.

[0023] In summary, the present application includes at least one of the following beneficial technical effects: filtering out web page traffic and application traffic from network traffic, and then determining the web page to be evaluated from which the web page traffic comes, and at the same time determining the application to be evaluated from which the application traffic comes. Further, the target fraud platform is screened out from each web page to be evaluated and each application to be evaluated, and then the corresponding target user is matched with an adapted target tag through the user behavior characteristics of the target user in the target fraud platform, and the first risk level of the target user as a fraud-related group is determined according to the target tag, so as to accurately evaluate the possibility that the target user is a fraud-related group. If the first risk level is greater than the level threshold, it means that the corresponding target user is likely to be a fraud-related group, and then it is determined as a suspected fraud-related user, so as to achieve accurate identification of the fraud-related group. Finally, according to the first abnormal network fingerprint, the traffic trend of the suspected fraud-related user is tracked, so as to facilitate the subsequent high-risk behavior of the fraud-related group, and can also be quickly and accurately identified. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a flowchart of a fraud-related group identification method provided in an embodiment of the present application; Figure 2 It is a flowchart of another method for identifying a fraudulent group provided in an embodiment of the present application; Figure 3 It is a structural schematic diagram of a fraud-related group identification device provided in an embodiment of the present application; Figure 4 It is a structural schematic diagram of another fraud-related group identification device provided in an embodiment of the present application.

[0025] Explanation of the accompanying drawings: 11. Traffic capture module; 12. Traffic screening module; 13. Behavior extraction module; 14. Label matching module; 15. Risk determination module; 16. Group identification module; 17. Anti-fraud reminder module. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments.

[0027] In the description of the embodiments of the present application, words such as "illustrative", "for example" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "illustrative", "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "illustrative", "for example" or "for example" is intended to present related concepts in a concrete way.

[0028] In the description of the embodiments of the present application, the term "and / or" is only a kind of association relationship describing the associated objects, indicating that there may be three kinds of relationships, for example, A and / or B, which can represent: A exists alone, B exists alone, and A and B exist at the same time. In addition, unless otherwise specified, the meaning of the term "multiple" refers to two or more. For example, multiple systems refer to two or more systems, and multiple screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Thus, the features defined as "first" and "second" can explicitly or implicitly include one or more of the features. The terms "include", "comprise", "have" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0029] See also Figure 1 The present application embodiment discloses a flow chart of a method for identifying a group of people involved in fraud, which can be implemented by a computer program or run on a group of people involved in fraud identification device based on the von Neumann system. The computer program can be integrated into an application or run as an independent tool application, specifically including: S101: Obtain network traffic.

[0030] Specifically, network traffic refers to the total amount of data transmitted in the network, which includes the data of each user in the network. User activities in the network will generate corresponding traffic. For example, users' activities such as web browsing, file downloading, video watching, page operation, etc. will generate corresponding traffic. These traffics are aggregated to form network traffic. Network traffic is mainly composed of application traffic and web traffic. Application traffic refers to the total amount of data transmission generated when users access Internet services through mobile applications (‌Application, app). Web traffic refers to the total amount of data transmission generated when users access Internet services through web browsers. Among them, traffic analysis of network traffic can provide insights into user behavior.

[0031] In the embodiment of the present application, the network traffic obtained is mainly the network traffic within a preset fixed time, which can be 10 minutes or 5 minutes. In other embodiments, the network traffic obtained can also be the network traffic captured in real time. Further, a feasible way to obtain network traffic is to capture the network traffic within a fixed time through a network device router or switch. In other embodiments, the network traffic can also be captured through a preset network monitoring tool, which is a Wireshark tool or a Tcpdump tool. In addition, the captured network traffic is stored in the cache area in the form of data packets.

[0032] It should be noted that the execution subject of a method for identifying a group involved in fraud is a server, and a client for identifying a group involved in fraud is installed in the terminal. After the terminal obtains network traffic through a router or switch, the network traffic is sent to the server through the client. The server is the backend server of the client, which can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. The terminal is a personal computer. In other embodiments, the execution subject of a method for identifying a group involved in fraud can also be a terminal directly.

[0033] S102: Filtering web page traffic and application traffic from the network traffic, and determining at least one web page to be evaluated corresponding to the web page traffic, and determining at least one application to be evaluated corresponding to the application traffic.

[0034] Specifically, after the network traffic is captured, it is necessary to filter out the web page traffic and application traffic. One feasible implementation method is: preprocessing the network traffic in the cache area. The preprocessing mainly includes deduplication, filtering and reorganization operations. Specifically, duplicate data packets in the network traffic are removed through data structures such as hash tables and Bloom filters, and irrelevant or invalid data packets in the network traffic are filtered out through regular expressions. The scattered data packets are reorganized through technologies such as TCP / IP protocol stack and session tracking to finally obtain the processed network traffic.

[0035] Furthermore, application layer data is extracted from the processed network traffic through Deep Packet Inspection (DPI), wherein Deep Packet Inspection is a network traffic analysis technology that deeply analyzes the content of data packets to identify and classify application layer protocols, services, and applications transmitted over the network. Among them, application layer data refers to transmission data located at the application layer during network communication. The application layer data contains traffic feature information in multiple dimensions. In an embodiment of the present application, the application layer data includes traffic feature data in dimensions such as request header User-Agent, Cookie, Referer, and Authorization. The request header is one of the important components of the application layer data and can be used to determine the source of the traffic. The application layer data also includes traffic feature information in dimensions such as access frequency, duration, and content characteristics. It should be noted that traffic feature information refers to the specific information corresponding to a single feature dimension in the application layer data.

[0036] Further, the first weight and the second weight of each flow characteristic information are determined, the first weight is the characteristic weight of the corresponding flow characteristic information in the web page, and the second weight is the characteristic weight of the corresponding flow characteristic information in the application, both of which are set based on the experience of the technicians. Exemplarily, taking the dimension of Cookie as an example, Web browsers usually carry and manage Cookie information, and apps usually carry a small amount of encrypted and verified Cookie information in API requests, then the characteristic weight of the flow characteristic information corresponding to Cookie in the web page is higher than the characteristic weight in the application, and then the corresponding first weight is greater than the second weight. By analogy, the first weight and the second weight corresponding to each flow characteristic information can be determined respectively. Further, the first score and the second score of each flow characteristic information are determined, the first score is the possibility of determining as web browsing based on the corresponding flow characteristic information, and the second score is the possibility of determining as application traffic based on the corresponding flow characteristic information. The first score and the second score both take values ​​from 0 to 1, the larger the value, the higher the possibility, and are both set based on human experience and input through the terminal. In addition, the sum of each first weight is 1, and the sum of each second weight is 1. Finally, the first scores of each traffic feature information are weighted and summed through the first weight to obtain the first score. Similarly, the second scores of each traffic feature information are weighted and summed through the second weight to obtain the second score. If the first score is greater than the second score, it means that the traffic is more likely to come from the web page, and then the corresponding part of the traffic is determined to be web page traffic; if the second score is greater than the first score, it means that the traffic is more likely to come from the application, and then the corresponding part of the traffic is determined to be application traffic, thereby realizing the screening of web page traffic and application traffic in network traffic.

[0037] Furthermore, the corresponding page to be evaluated is determined through the screened web traffic, that is, the page from which the web traffic comes, and the corresponding page to be evaluated is specifically crawled through a preset Web Scraping tool. Since the web traffic may include web traffic from different web pages, there is at least one page to be evaluated. In addition, through the screened application traffic, the corresponding application to be evaluated is determined through a preset crawler, and the installation program file of the application to be evaluated is crawled.

[0038] S103: Determine a target fraud platform from each to-be-evaluated web page and each to-be-evaluated application, and extract user behavior characteristics of target users in the target fraud platform.

[0039] Specifically, after each web page to be evaluated and each application to be evaluated are determined, it is necessary to determine whether each web page to be evaluated is a fraudulent web page and whether each application to be evaluated is a fraudulent application. If yes, it is determined to be a target fraud platform. In an embodiment of the present application, a feasible way to determine whether a web page to be evaluated is a fraudulent web page is to parse and extract metadata features in the web page to be evaluated, where the metadata features include node tags such as javascript, css, image, and div, and then count the existence and quantity of these node tags. Based on the determined existence and quantity, the metadata features are converted into a vector representation as follows: [{javascript_count},{css_count},{image_count},{div_count},{javascript_exists},{css_exists}, {image_exists}, {div_exists}], where {javascript_count} indicates the number of javascript node tags on the page to be evaluated, {css_count} indicates the number of css node tags on the page to be evaluated, {image_count} indicates the number of image node tags on the page to be evaluated, {div_count} indicates the number of div node tags on the page to be evaluated, {javascript_exists} indicates whether the javascript node tag on the page to be evaluated exists, {css_exists} indicates whether the css node tag on the page to be evaluated exists, {image_exists} indicates whether the image node tag on the page to be evaluated exists, and {div_exists} indicates whether the div node tag on the page to be evaluated exists. It should be noted that existence is either existence or non-existence.

[0040] Furthermore, the cosine similarity between the vector corresponding to the metadata feature in the web page to be evaluated and each first reference vector in the preset first fraud-related library is calculated. The larger the cosine similarity, the more likely the corresponding web page to be evaluated is a fraudulent web page. If the cosine similarity exceeds the preset similarity threshold, the corresponding web page to be evaluated is determined to be a fraudulent web page, wherein the first fraud-related library includes vectors corresponding to metadata features in different fraudulent web pages, i.e., first reference vectors. In other embodiments, the Euclidean distance between the vector corresponding to the metadata feature in the web page to be evaluated and each first reference vector may also be calculated. The smaller the Euclidean distance, the more likely the corresponding web page to be evaluated is a fraudulent web page.

[0041] Furthermore, a feasible way to determine whether the application to be evaluated is a fraudulent application is: parse the installation program file of the application to be evaluated to obtain the corresponding metadata features, specifically through the preset APK Analyser tool for analysis, the metadata features corresponding to the application to be evaluated include application name, package name, file size, MD5, version number, certificate hash value, certificate SHA-1 value‌ and developer information, etc., and convert these metadata features into vector representation, and then calculate the cosine similarity between the vector corresponding to the metadata feature in the application to be evaluated and each second reference vector in the preset second fraud-related library. The larger the cosine similarity, the more likely the corresponding application to be evaluated is a fraudulent application. If the cosine similarity exceeds the preset similarity threshold, it is determined that the corresponding application to be evaluated is a fraudulent application, wherein the second fraud-related library includes vectors corresponding to metadata features in different fraudulent applications, i.e., second reference vectors.

[0042] After the target fraud platform is determined, it is necessary to extract the user behavior characteristics of the target users on the target fraud platform, that is, to extract the corresponding user behavior characteristics from the user traffic of the target users in the network traffic. The target user is the user who accesses and operates the target fraud platform. In the embodiment of the present application, a feasible extraction method is: using protocol analysis technology to identify the application layer protocol or service used in the network traffic corresponding to the target fraud platform, and extracting the user behavior characteristics of the target user based on the content or parameters carried in the application layer protocol or service. Among them, protocol analysis technology is a process of obtaining data information by decoding the header and tail of the network protocol, which is used to identify application layer data in network traffic.

[0043] S104: Based on at least one user behavior feature of the same target user, determine a target tag of the corresponding target user from a preset user tag library, where the user tag library includes at least one tag possessed by a fraud-related group.

[0044] Specifically, after the user behavior characteristics of the target user are extracted, for a single target user, a target tag is matched for the corresponding target user from a preset user tag library according to the corresponding user behavior characteristics. In an embodiment of the present application, the user tag library includes at least one tag possessed by the fraud-related group. Exemplarily, the fraud-related group is a pig-killing fraud team, then the user tag library includes role tags such as the feeding group, the speech group, and the fund management group. In other embodiments, behavior tags may also be included. For example, the fraud-related group adds and removes fake products, modifies virtual stocks, and so on.

[0045] Furthermore, a feasible way to determine the target tag is to calculate the abnormal score value of the user behavior characteristics of the same target user through a preset abnormal score value calculation formula, wherein the abnormal score value calculation formula is: s=W f ·( 100)+W t ·( 100) + W i ·( 100)+W p ·( 100)+W a ·( 100); Where W f , W t , W i , W p and W a All represent fixed weights, the sum of all fixed weights is 1, f observed The behavior frequency representing the user behavior characteristics in the target fraud platform, f threshold represents the preset behavior frequency threshold, t observed represents the duration of user behavior characteristics in the target fraud platform, t threshold represents the preset duration threshold, i observed represents the number of interactions of the target user in the target fraud platform, i threshold represents the preset interaction number threshold, p observed represents the payment amount of the target user on the target fraud platform, p threshold Indicates the preset payment amount threshold, a observed Indicates the number of occurrences of the same type of behavior characteristics as the user behavior characteristics, a threshold represents the preset occurrence threshold, and s represents the abnormal score value of the user behavior feature. It should be noted that the above f observed ,t observed 、i observed 、p observed and a observedAll of them can be obtained from the traffic log corresponding to the network traffic. The feasible way to determine the traffic log is: determine the corresponding process log through the preset Zabbix tool. In other embodiments, it can also be determined through the preset Nagios tool. In addition, the number of interactions in the target fraud platform refers to the number of browsing, commenting, etc. in the target fraud platform. Behavior characteristics of the same type as user behavior characteristics can be matched from a preset behavior matching table of the same type according to user behavior characteristics. The behavior matching table of the same type includes different user behavior characteristics and corresponding behavior characteristics of the same type. Exemplarily, the behavior matching table of the same type may include: user behavior characteristics are browsing products, and the behavior characteristics of the same type are collecting products, both involving e-commerce shopping; user behavior characteristics are browsing investment products, and the behavior characteristics of the same type are purchasing virtual digital products, both involving investment and financial management.

[0046] For example, if the user behavior feature m of the target user M in the target fraud platform is “fraud platform operation”, then the number of times the target user M opens the target fraud platform is 50, that is, f observed is 50 times, and the preset behavior frequency threshold is 30 times; the target user M stays on this target fraud platform for 1000 minutes, that is, t observed is 1000 minutes, and the preset duration threshold is 500 minutes; the number of interactions of the target user M on the target fraud platform is 70 times, that is, i observed is 70 times, and the preset interaction threshold is 40 times; the target user M pays 2,000 yuan on this target fraud platform, that is, p observed is 2000 yuan, and the preset payment amount threshold is 1000 yuan; the number of times the same type of behavior feature as user behavior feature m appears is 10 times, that is, a observed is 10 times, and the preset occurrence threshold is 5 times. Each fixed weight is set to 0.2, and finally each parameter is substituted into the abnormal score calculation formula, which is: 0.2·(·100)+ 0.2·(·100)+ 0.2·(·100)+ 0.2·(·100)+ 0.2·(·100)=188.33, and the abnormal score value of user behavior feature m is 188.33. It should be noted that the higher the abnormal score value, the higher the abnormality of the user behavior feature of the target user.

[0047] Furthermore, the corresponding abnormal score values ​​of other user behavior characteristics of the same target user can also be calculated in the same way, and then the various abnormal score values ​​corresponding to the same target user are constructed into a matrix to obtain the abnormal behavior feature vector of this target user on the target fraud platform. Furthermore, according to each abnormal score value in the abnormal behavior feature vector, the corresponding adaptation label is determined from the preset user label library. In addition to the different labels possessed by the fraud-related groups, the user label library also includes the user behavior characteristics and abnormal score values ​​corresponding to the label. Finally, each adaptation label is determined as the target label of the corresponding target user, that is, the label of the fraud-related group matched by the target user.

[0048] S105: Based on the target tag corresponding to the same target user, calculate the first risk level of the corresponding target user being a fraud-related group. The higher the first risk level, the higher the possibility that the corresponding target user is a fraud-related group.

[0049] Specifically, after the target label corresponding to the same target user is determined, when each target label is a label possessed by a fraud-related group, the first risk level of the corresponding target user as a fraud-related group is calculated. The higher the first risk level, the higher the possibility that the corresponding target user is a fraud-related group. In an embodiment of the present application, a feasible way to calculate the first risk level is to determine each abnormal score value of the same target user as the label value of the corresponding target label, and then perform weighted summation on each label value to obtain the corresponding target user's fraud-related risk score. Further, this fraud-related risk score is substituted into the preset risk level calculation formula to obtain the first risk level of the corresponding target user as a fraud-related group. Among them, the risk level calculation formula is: ; In the formula, R represents the first risk level, S represents the fraud risk score, and P represents the possibility of the target user committing fraud, ranging from 0 to 1. Indicates the number of times the target user has triggered the risk within the latest preset time. It indicates the total number of times the target user has triggered risks in history, and T indicates the duration corresponding to the preset time. and All represent adjustment parameters. Represents the scaling parameter.

[0050] It should be noted that triggering risk refers to the situation where the target user's previously determined risk level is higher than the preset level threshold. In the embodiment of the present application, it refers to the situation where the target user's previously determined first risk level is higher than the preset level threshold. The product of S and P reflects the probability of the target user currently committing fraud. The ratio to T is used to measure the frequency of recent high-risk behaviors of target users. The effect is to reduce The impact of being too small on calculating the first risk level, Reflects the cumulative effect of the target users’ long-term high-risk behaviors. Taking the logarithm can effectively avoid over-amplifying the impact of historical high-risk behaviors and ensure the accuracy of the first risk level of the current target user. In addition, The effect is to reduce The impact of being too small on the calculation of the first risk level. α is used to adjust the impact of recent high-risk behaviors on the first risk level. The higher the value, the greater the impact of recent high-risk behaviors. γ is used to adjust the impact of historical high-risk behaviors on the first risk level. The higher the value, the greater the impact of historical high-risk behaviors. β is used to scale the impact of the total number of historical risk triggers. In addition, a feasible way to determine the possibility of a target user committing fraud is to input each user behavior feature of the target user into a preset prediction model to determine the possibility of the target user committing fraud, wherein the prediction model can use a binary classification neural network model or a recurrent neural network model, and the training process can be briefly described as: the user behavior features marked with "committed fraud" and "did not commit fraud" are used as training to train the model until the model converges, and finally a prediction model is obtained. This is a prior art and will not be repeated here.

[0051] S106: When the first risk level is greater than a preset level threshold, the corresponding target user is determined as a suspected fraud user, and the corresponding first abnormal network fingerprint is determined based on the user traffic of the suspected fraud user. Based on the first abnormal network fingerprint, the traffic trend of the corresponding suspected fraud user is tracked.

[0052] Specifically, the first risk level is compared with the preset level threshold. If the first risk level is greater than the level threshold, it means that the corresponding target user is more likely to be a fraud-related group, and is determined as a suspected fraud-related user. Further, the user flow of the suspected fraud-related user is extracted from the network flow through the preset flowcontainer library, and the corresponding first abnormal network fingerprint is determined based on the user flow. In the embodiment of the present application, the feasible way to determine the first abnormal network fingerprint is to extract at least one device feature information from the user flow of the suspected fraud-related user through the preset Wireshark tool, which is specifically divided into six categories: basic fingerprint information, advanced fingerprint information, geographic location information, time zone information, DNS information, and SSL certificate information. Among them, basic fingerprint information includes User-agent, browser version, mobile phone model, memory size, plug-in information, language version, screen information, etc.; advanced fingerprint information includes Canvas fingerprint, WebGL fingerprint, AudioContext fingerprint, WebRTC fingerprint, etc.; geographic location information includes IP address, latitude and longitude, WIFI hotspot, base station information, etc.; time zone information includes system time zone, local time, network time, etc.; DNS information includes domain name resolution server, resolution result, resolution time, etc.; SSL certificate information includes certificate authority, certificate validity period, certificate fingerprint, etc. Further, the feature types of each device feature information are determined, which are: basic fingerprint information, advanced fingerprint information, geographic location information, time zone information, DNS information and SSL certificate information. According to different feature types, each device feature information is divided into multiple levels, namely, the basic layer, advanced layer, geographical layer, time layer and network layer. The device feature information of the feature type of basic fingerprint information is classified as the basic layer, the device feature information of the feature type of advanced fingerprint information is classified as the advanced layer, the device feature information of the feature type of geographic location information is classified as the geographical layer, the device feature information of the feature type of time zone information is classified as the time layer, and the device feature information of the feature type of DNS information and SSL certificate information is classified as the network layer. It should be noted that in the embodiment of the present application, each device feature information is divided into five different levels in order to independently calculate and analyze device features at different levels, reduce the impact of device feature anomalies in a single level on the determination of network fingerprints, and improve the accuracy of the subsequent first abnormal network fingerprint. In addition, a network fingerprint refers to a unique device identifier generated through network activities, which is used to identify and track network devices.

[0053] Furthermore, the device feature information in each layer is aggregated. One feasible aggregation calculation method is to perform weighted summation on the device feature information in each layer to obtain the corresponding intra-layer aggregation result. In other embodiments, the feature information of each device can also be aggregated by principal component analysis.

[0054] After the intra-layer aggregation results corresponding to each level are determined, the weighted sum of the intra-layer aggregation results is performed to obtain the inter-layer combination results, and finally the inter-layer combination results are hashed by a preset hash algorithm to obtain the first abnormal network fingerprint corresponding to the suspected fraudulent user. Among them, the hash algorithm can use MD4 or MD5. In other embodiments, the hash algorithm can also be used to hash each intra-layer aggregation result separately to obtain multiple hash results, and finally the hash results are combined together to obtain a composite hash value and determine it as the first abnormal network fingerprint. Furthermore, through the preset FingerprintJS, based on the first abnormal network fingerprint, the traffic trend of the corresponding suspected fraudulent user is tracked, that is, the network activity of the device of the suspected fraudulent user is tracked, and the occurrence of fraud cases is curbed to a certain extent. Among them, the first abnormal network fingerprint is a unique identifier of the suspected fraudulent user device.

[0055] See also Figure 2 , the present application embodiment discloses a flow chart of another method for identifying a group of people involved in fraud, which can be implemented by a computer program or run on a group of people involved in fraud identification device based on the von Neumann system. The computer program can be integrated into an application or run as an independent tool application, specifically including: S201: Obtain network traffic.

[0056] S202: Filtering web page traffic and application traffic from network traffic, and determining at least one web page to be evaluated corresponding to the web page traffic, and determining at least one application to be evaluated corresponding to the application traffic.

[0057] S203: Determine a target fraud platform from each to-be-evaluated web page and each to-be-evaluated application, and extract user behavior characteristics of target users in the target fraud platform.

[0058] S204: Based on at least one user behavior feature of the same target user, determine a target tag of the corresponding target user from a preset user tag library, where the user tag library includes at least one tag possessed by a fraud-related group and at least one tag possessed by a fraud-victim group.

[0059] S205: Based on the target tag corresponding to the same target user, calculate the first risk level of the corresponding target user being a fraud-related group. The higher the first risk level, the higher the possibility that the corresponding target user is a fraud-related group.

[0060] S206: When the first risk level is greater than a preset level threshold, the corresponding target user is determined as a suspected fraud user, and the corresponding first abnormal network fingerprint is determined based on the user traffic of the suspected fraud user, and the traffic trend of the corresponding suspected fraud user is tracked based on the first abnormal network fingerprint.

[0061] For details, please refer to steps S101-S106, which will not be described in detail here.

[0062] S207: When the target tags corresponding to the same target user are all tags possessed by the fraud-victims group, the second risk level of the corresponding target user as the fraud-victims group is calculated based on each target tag.

[0063] S208: When the second risk level is greater than a preset level threshold, the corresponding target user is determined as a suspected defrauded user, and a corresponding second abnormal network fingerprint is determined according to the user traffic of the suspected defrauded user.

[0064] S209: Determine whether the corresponding suspected defrauded user is accessing a fraud platform based on the second abnormal network fingerprint, and if so, send anti-fraud reminder information to the device of the corresponding suspected defrauded user.

[0065] Specifically, after determining the target tag corresponding to the target user, if the target tags corresponding to the same target user are all tags possessed by the fraud group, it means that this target user is most likely not a fraud group. Then, according to each target tag, calculate the second risk level of the corresponding target user as a fraud group, thereby evaluating the risk of the target user as a fraud group. For the specific calculation process, refer to the part of step S105 in which the first risk level of the corresponding target user as a fraud group is calculated according to the target tag. Perform a weighted sum on the label values ​​of each target label to obtain a fraud risk score, and substitute it into the fraud risk level calculation formula: Finally, the corresponding second risk level is obtained.

[0066] In the formula, R' represents the second risk level, S' represents the fraud risk score, and P' represents the possibility of the target user being defrauded, ranging from 0 to 1. Indicates the number of times the target user has triggered the risk within the latest preset time. Indicates the total number of times the target user has triggered risks in history, T` indicates the duration corresponding to the preset time, and All represent adjustment parameters. In the embodiment of the present application, the trigger risk refers to the situation that the second risk level determined by the target user in the past is higher than the preset level threshold.

[0067] If the second risk level is greater than the preset level threshold, it means that the corresponding target user is likely to be a defrauded group, so the corresponding target user is determined as a suspected defrauded user. Further, the user traffic of the suspected defrauded user is extracted from the network traffic, and the corresponding second abnormal network fingerprint is determined. The determination process can be found in step S106 and will not be repeated here.

[0068] Finally, based on the second abnormal network fingerprint, the traffic trend of the corresponding suspected fraud group can be tracked to determine whether they are accessing a fraud platform. If the platform they are accessing is a fraud platform, anti-fraud reminder information is sent to the device of the corresponding suspected fraud user, thereby curbing the occurrence of fraud cases to a certain extent. It should be noted that the second abnormal network fingerprint is a unique identifier of the device of the suspected fraud user. In addition, the device can be a smartphone or a personal computer.

[0069] The implementation principle of a method for identifying a fraudulent group in an embodiment of the present application is as follows: web page traffic and application traffic are screened out from network traffic, and then the web page to be evaluated from which the web page traffic originates is determined, and at the same time, the application to be evaluated from which the application traffic originates is determined. Further, the target fraud platform is screened out from each web page to be evaluated and each application to be evaluated, and then the corresponding target user is matched with an adapted target tag through the user behavior characteristics of the target user in the target fraud platform, and the first risk level of the target user as a fraudulent group is determined based on the target tag, so as to accurately evaluate the possibility that the target user is a fraudulent group. If the first risk level is greater than the level threshold, it means that the corresponding target user is likely to be a fraudulent group, and then it is determined to be a suspected fraudulent user, so as to achieve accurate identification of the fraudulent group. Finally, according to the first abnormal network fingerprint, the traffic trend of the suspected fraudulent user is tracked, so as to facilitate the subsequent high-risk behavior of the fraudulent group, and also to quickly and accurately identify it.

[0070] The following is an embodiment of the device of the present application, which can be used to execute the embodiment of the method of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the embodiment of the method of the present application.

[0071] See also Figure 3 , which is a schematic diagram of the structure of the fraud group identification device provided in the embodiment of the present application. The fraud group identification device can be implemented as all or part of the device through software, hardware or a combination of both. The device includes a traffic capture module 11, a traffic screening module 12, a behavior extraction module 13, a label matching module 14, a risk determination module 15 and a group identification module 16.

[0072] Traffic capture module 11, used to obtain network traffic; The traffic screening module 12 is used to screen web traffic and application traffic from the network traffic, and determine at least one web page to be evaluated corresponding to the web traffic, and determine at least one application to be evaluated corresponding to the application traffic; The behavior extraction module 13 is used to determine the target fraud platform from each to-be-evaluated web page and each to-be-evaluated application, and extract the user behavior characteristics of the target user in the target fraud platform, where the target user is the user who accesses and operates the target fraud platform; The tag matching module 14 is used to determine the target tag of the corresponding target user from a preset user tag library based on at least one user behavior feature of the same target user, where the user tag library includes at least one tag possessed by the fraud-related group; The risk determination module 15 is used to calculate the first risk level of the corresponding target user as a fraud-related group based on the target tag corresponding to the same target user. The higher the first risk level, the higher the possibility that the corresponding target user is a fraud-related group; The group identification module 16 is used to determine the corresponding target user as a suspected fraud user when the first risk level is greater than a preset level threshold, determine the corresponding first abnormal network fingerprint based on the user traffic of the suspected fraud user, and track the traffic trend of the corresponding suspected fraud user based on the first abnormal network fingerprint.

[0073] Optionally, the tag matching module 14 is specifically used for: Calculate the abnormal score value of the user behavior characteristics of the same target user through a preset abnormal score value calculation formula; Based on the abnormal score values ​​corresponding to the same target user, the abnormal behavior feature vector of the corresponding target user on the target fraud platform is obtained; According to each abnormal score value in the abnormal behavior feature vector, a corresponding adaptation tag is matched from a preset user tag library, and each adaptation tag is determined as a target tag of the corresponding target user.

[0074] Optionally, the risk determination module 15 is specifically used for: Determine each abnormal score value of the same target user as the label value of the corresponding target label, and perform weighted summation on the label values ​​to obtain the corresponding target user's fraud risk score; Substitute the fraud risk score into the preset risk level calculation formula to obtain the first risk level of the corresponding target user as a fraud group. The risk level calculation formula is: ; In the formula, R represents the first risk level, S represents the fraud risk score, and P represents the possibility of the target user committing fraud, ranging from 0 to 1. Indicates the number of times the target user has triggered the risk within the latest preset time. It indicates the total number of times the target user has triggered risks in history, and T indicates the duration corresponding to the preset time. and All represent adjustment parameters. Represents the scaling parameter.

[0075] Optionally, the group identification module 16 is specifically used for: Extract at least one device feature information from the user traffic of the suspected fraudulent user; Determine the characteristic type of each device characteristic information, and divide the each device characteristic information into multiple levels according to the characteristic type; Aggregate and calculate the characteristic information of each device in each layer to obtain the corresponding intra-layer aggregation results; The aggregation results in each layer are weighted and summed to obtain the inter-layer combination results, and the inter-layer combination results are hashed using a preset hash algorithm to obtain the first abnormal network fingerprint corresponding to the suspected fraudulent user.

[0076] Optional, such as Figure 4 As shown, the device also includes an anti-fraud reminder module 17, which is specifically used to: When the target tags corresponding to the same target user are all tags possessed by the fraud-victims group, the second risk level of the corresponding target user as the fraud-victims group is calculated based on each target tag; When the second risk level is greater than a preset level threshold, the corresponding target user is determined as a suspected defrauded user, and a corresponding second abnormal network fingerprint is determined based on the user traffic of the suspected defrauded user; According to the second abnormal network fingerprint, it is determined whether the corresponding suspected defrauded user is accessing the fraud platform. If so, an anti-fraud reminder message is sent to the device of the corresponding suspected defrauded user.

[0077] Optionally, the traffic screening module 12 is specifically used for: Preprocess the network traffic to obtain processed network traffic; Extracting application layer data from the processed network traffic through deep packet inspection, and determining a first weight and a second weight of each flow feature information in the application layer data, wherein the first weight is a feature weight of the corresponding flow feature information in the web page, and the second weight is a feature weight of the corresponding flow feature information in the application; According to the first weight, weighted sum of the first scores of each flow characteristic information is performed to obtain a first score, and according to the second weight, weighted sum of the second scores of each flow characteristic information is performed to obtain a second score; If the first score is greater than the second score, the corresponding part of the traffic is determined as web page traffic; if the first score is less than the second score, the corresponding part of the traffic is determined as application traffic.

[0078] It should be noted that the above embodiment provides a group identification device for fraudulent groups. When executing the group identification method for fraudulent groups, only the division of the above functional modules is used as an example. In actual applications, the above functional distribution can be completed by different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the group identification device for fraudulent groups provided in the above embodiment and the method embodiment for group identification for fraudulent groups belong to the same concept. The implementation process is detailed in the method embodiment and will not be repeated here.

[0079] The embodiment of the present application further discloses a computer-readable storage medium, and the computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, a method for identifying a fraud-related group according to the above embodiment is adopted.

[0080] Among them, the computer program can be stored in a computer-readable medium, the computer program includes computer program code, the computer program code can be in the form of source code, object code, executable file or certain middleware, etc. The computer-readable medium includes any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the computer-readable medium includes but is not limited to the above-mentioned components.

[0081] Among them, through this computer-readable storage medium, a fraud-related group identification method of the above embodiment is stored in a computer-readable storage medium, and is loaded and executed on a processor to facilitate the storage and application of the above method.

[0082] An embodiment of the present application further discloses an electronic device, in which a computer program is stored in a computer-readable storage medium. When the computer program is loaded and executed by a processor, the above-mentioned method for identifying a fraudulent group is adopted.

[0083] The electronic device may be a desktop computer, a laptop computer, a cloud server or other electronic device, and the electronic device includes but is not limited to a processor and a memory. For example, the electronic device may also include input and output devices, a network access device, and a bus.

[0084] Among them, the processor can adopt a central processing unit (CPU). Of course, according to actual usage, other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. can also be adopted. The general-purpose processor can adopt a microprocessor or any conventional processor, etc., and this application does not impose any restrictions on this.

[0085] Among them, the memory can be an internal storage unit of the electronic device, such as a hard disk or memory of the electronic device, or it can be an external storage device of the electronic device, such as a plug-in hard disk, a smart memory card (SMC), a secure digital card (SD) or a flash memory card (FC) equipped on the electronic device. Moreover, the memory can also be a combination of an internal storage unit and an external storage device of the electronic device. The memory is used to store computer programs and other programs and data required by the electronic device. The memory can also be used to temporarily store data that has been output or is to be output, and this application does not impose any restrictions on this.

[0086] Among them, through this electronic device, a fraud-related group identification method of the above embodiment is stored in the memory of the electronic device, and is loaded and executed on the processor of the electronic device for easy use.

[0087] The above is only an exemplary embodiment of the present disclosure, and the scope of the present disclosure cannot be limited thereto. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the specification and practicing the disclosure here, those skilled in the art will easily think of other embodiments of the present disclosure. This application is intended to cover any modification, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the technical field not recorded in the present disclosure. The description and examples are only regarded as exemplary, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A method for identifying a fraudulent group, characterized in that: The method comprises: Get network traffic; Filtering web page traffic and application traffic from the network traffic, and determining at least one web page to be evaluated corresponding to the web page traffic, and determining at least one application to be evaluated corresponding to the application traffic; Determine a target fraud platform from each of the web pages to be evaluated and each of the applications to be evaluated, and extract user behavior characteristics of a target user in the target fraud platform, wherein the target user is a user who accesses and operates the target fraud platform; Based on at least one user behavior feature of the same target user, determine a target tag of the corresponding target user from a preset user tag library, wherein the user tag library includes at least one tag possessed by a fraud-related group; Based on the target tag corresponding to the same target user, calculating the first risk level of the corresponding target user as a fraud-related group, wherein the higher the first risk level, the higher the possibility that the corresponding target user is a fraud-related group; When the first risk level is greater than a preset level threshold, the corresponding target user is determined as a suspected fraud user, and the corresponding first abnormal network fingerprint is determined based on the user traffic of the suspected fraud user. Based on the first abnormal network fingerprint, the traffic trend of the corresponding suspected fraud user is tracked.

2. The fraud-related group identification method according to claim 1 is characterized in that: The determining of the target tag of the corresponding target user from a preset user tag library based on at least one user behavior feature of the same target user specifically includes: Calculate the abnormal score value of the user behavior characteristics of the same target user through a preset abnormal score value calculation formula; Based on each abnormal score value corresponding to the same target user, obtain an abnormal behavior feature vector of the corresponding target user on the target fraud platform; According to each abnormal score value in the abnormal behavior feature vector, a corresponding adaptation tag is matched from a preset user tag library, and each adaptation tag is determined as a target tag of the corresponding target user.

3. The fraud-related group identification method according to claim 2 is characterized in that: The calculation formula of the abnormal score value is: s=W f ·( ·100)+ W t ·( ·100) + W i ·( ·100)+ W p ·( ·100)+ W a ·( ·100); Where W f , W t , W i , W p and W a All represent fixed weights, the sum of all fixed weights is 1, f observed The behavior frequency representing the user behavior characteristics in the target fraud platform, f threshold represents the preset behavior frequency threshold, t observed represents the duration of user behavior characteristics in the target fraud platform, t threshold represents the preset duration threshold, i observed represents the number of interactions of the target user in the target fraud platform, i threshold represents the preset interaction number threshold, p observed represents the payment amount of the target user on the target fraud platform, p threshold Indicates the preset payment amount threshold, a observed Indicates the number of occurrences of the same type of behavior characteristics as the user behavior characteristics, a threshold represents the preset occurrence threshold, and s represents the abnormal score value of the user behavior feature.

4. The fraud-related group identification method according to claim 2 is characterized in that: The step of calculating the first risk level of the corresponding target user as a fraud-related group based on the target tag corresponding to the same target user specifically includes: Determine each abnormal score value of the same target user as the label value of the corresponding target label, and perform weighted summation on the label values ​​to obtain the corresponding target user's fraud risk score; Substitute the fraud risk score into the preset risk level calculation formula to obtain the first risk level of the corresponding target user as a fraud group. The risk level calculation formula is: ; In the formula, R represents the first risk level, S represents the fraud risk score, and P represents the possibility of the target user committing fraud, ranging from 0 to 1. Indicates the number of times the target user has triggered the risk within the latest preset time. It indicates the total number of times the target user has triggered risks in history, and T indicates the duration corresponding to the preset time. and All represent adjustment parameters. Represents the scaling parameter.

5. The fraud-related group identification method according to claim 1 is characterized in that: The determining, according to the user traffic of the suspected fraudulent user, a corresponding first abnormal network fingerprint specifically includes: Extracting at least one device feature information from the user traffic of the suspected fraudulent user; Determining a feature type of each device feature information, and dividing each device feature information into multiple levels according to the feature type; Aggregate and calculate the characteristic information of each device in each layer to obtain the corresponding intra-layer aggregation result; The aggregation results within each layer are weightedly summed to obtain an inter-layer combination result, and the inter-layer combination result is hashed using a preset hash algorithm to obtain a first abnormal network fingerprint corresponding to the suspected fraud-related user.

6. The fraud-related group identification method according to claim 1, characterized in that: The user tag library includes at least one tag possessed by the defrauded group, and the method further includes: When the target tags corresponding to the same target user are all tags possessed by the fraud-stricken group, the second risk level of the corresponding target user as the fraud-stricken group is calculated according to each of the target tags; When the second risk level is greater than a preset level threshold, the corresponding target user is determined as a suspected defrauded user, and a corresponding second abnormal network fingerprint is determined according to the user traffic of the suspected defrauded user; According to the second abnormal network fingerprint, it is determined whether the corresponding suspected defrauded user is accessing a fraud platform. If so, an anti-fraud reminder message is sent to the device of the corresponding suspected defrauded user.

7. The fraud-related group identification method according to claim 1, characterized in that: The filtering of web page traffic and application traffic from the network traffic specifically includes: Preprocessing the network traffic to obtain processed network traffic; Extracting application layer data from the processed network traffic through deep packet inspection, and determining a first weight and a second weight of each flow feature information in the application layer data, wherein the first weight is a characteristic weight of the corresponding flow feature information in the webpage, and the second weight is a characteristic weight of the corresponding flow feature information in the application; According to the first weight, weighted sum of the first scores of each piece of traffic feature information is performed to obtain a first score, and according to the second weight, weighted sum of the second scores of each piece of traffic feature information is performed to obtain a second score; If the first score is greater than the second score, the corresponding part of the traffic is determined as web page traffic; if the first score is less than the second score, the corresponding part of the traffic is determined as application traffic.

8. A device for identifying fraudulent groups, characterized in that: include: A traffic capture module (11), used for acquiring network traffic; A traffic screening module (12) is used to screen web page traffic and application traffic from the network traffic, and determine at least one web page to be evaluated corresponding to the web page traffic, and determine at least one application to be evaluated corresponding to the application traffic; A behavior extraction module (13) is used to determine a target fraud platform from each of the web pages to be evaluated and each of the applications to be evaluated, and to extract user behavior characteristics of a target user in the target fraud platform, wherein the target user is a user who accesses and operates the target fraud platform; A tag matching module (14) is used to determine a target tag of a corresponding target user from a preset user tag library based on at least one user behavior feature of the same target user, wherein the user tag library includes at least one tag possessed by a fraud-related group; A risk determination module (15) is used to calculate, based on the target tag corresponding to the same target user, a first risk level of the corresponding target user being a fraud-related group, wherein the higher the first risk level, the higher the possibility that the corresponding target user is a fraud-related group; The group identification module (16) is used to determine the corresponding target user as a suspected fraud-related user when the first risk level is greater than a preset level threshold, determine the corresponding first abnormal network fingerprint based on the user traffic of the suspected fraud-related user, and track the traffic trend of the corresponding suspected fraud-related user based on the first abnormal network fingerprint.

9. A computer-readable storage medium having a computer program stored therein, characterized in that: When the computer program is loaded and executed by a processor, the method according to any one of claims 1 to 7 is adopted.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that: When the processor loads and executes the computer program, the method according to any one of claims 1 to 7 is adopted.

Citation Information

Patent Citations

  • Determining traffic quality using event-based traffic scoring

    CN107077455A

  • Fraudulent behavior identification method and device

    CN118052594A

  • Fraud identification method and device, electronic equipment and storage medium

    CN118674465A

Cited By

  • LSTM-based fraud-related network flow identification method and server

    CN119945789A