Account security assessment method, device and equipment based on neural network

Abstract

The application provides a neural network-based account security evaluation method, device and equipment, the method comprising: obtaining a plurality of account information of a target account and a plurality of article information of articles published by the target account in a preset time period; neural network training uses a neural network to determine an account risk index corresponding to the target account according to the plurality of account information of the target account, and determines an article publishing risk index corresponding to the target account according to the plurality of article information of the articles published by the target account in the preset time period; determining a network position content security index corresponding to the target account according to the account risk index and the article publishing risk index; and generating an evaluation result based on the network position content security index. In the embodiment of the application, the account information does not need to be analyzed manually, thereby improving the evaluation efficiency; a specific mathematical model and a statistical method do not need to be selected, and the application has universality; and the neural network training completes data processing through a security evaluation model, thereby improving the accuracy of the evaluation result.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Internet, and particularly relates to a neural network-based account security assessment method, device and equipment. BACKGROUND

[0002] With the rapid development of the Internet, emerging communication technologies and network positions such as microblogs, WeChat public accounts and various self-media APPs have emerged. These platforms have greatly facilitated people's lives, changed the production methods of traditional industries, and promoted the rapid development and transformation and upgrading of the national economy.

[0003] At present, the traditional method of security assessment of accounts mainly includes two categories: qualitative assessment and quantitative assessment.

[0004] In the process of qualitative assessment, experts will conduct in-depth analysis of the content source, type, transmission channel, audience characteristics and inherent potential malicious behavior patterns in the content. The reliability of the content source, the sensitivity of the type, the universality of the transmission channel, the diversity of the audience characteristics and the complexity of the potential malicious behavior patterns are all indispensable factors in the assessment. Through the comprehensive consideration of these factors, experts can more accurately grasp the risk points of content security. However, qualitative assessment also has certain limitations. Due to the high dependence on the subjective judgment and experience of experts, the evaluation results may be affected by the personal bias, knowledge level and other factors of the experts. In addition, qualitative assessment is usually difficult to realize the rapid assessment of a large amount of content, which may become a bottleneck when facing a large amount of network content. The qualitative assessment method highly depends on the rich practical experience and deep professional knowledge of network security experts. Through in-depth analysis of various parameters affecting content security, including content source, type, transmission channel, audience characteristics and inherent potential malicious behavior patterns in the content, the method comprehensively identifies the potential security risks in the content. Experts use various technologies such as text analysis, semantic understanding and sentiment detection in the assessment process, and combine the deep understanding of industry dynamics, policies and regulations to subjectively evaluate the security of the content. This process finally obtains the qualitative description and classification of the content security risk.

[0005] Quantitative evaluation is a scientific application of mathematical models and statistical methods. Quantitative evaluation can avoid subjective interference of human factors and obtain more objective and accurate evaluation results. However, quantitative evaluation also has certain limitations. First, the selection and application of mathematical models and statistical methods require certain professional knowledge and skills, which may be difficult for non-professionals. Second, quantitative evaluation can only deal with quantifiable and measurable risk factors, and may not be able to effectively evaluate risk factors that are difficult to quantify and measure. In addition, the results of quantitative evaluation may be affected by data quality and sample size, and if the data quality is not high or the sample size is insufficient, the evaluation results may be biased or inaccurate.

[0006] Therefore, how to provide an account security evaluation method with accurate evaluation results and high evaluation efficiency is a technical problem to be solved by those skilled in the art. SUMMARY

[0007] The main purpose of the present application is to provide a neural network-based account security evaluation method, device and equipment, which aims to solve the technical problems of inaccurate evaluation results and low evaluation efficiency of the existing account security evaluation methods.

[0008] To achieve the above-mentioned purpose, the present application provides a neural network-based account security evaluation method, which comprises:

[0009] Obtaining a plurality of account information of a target account and a plurality of article information of articles published by the target account in a preset time period;

[0010] According to the plurality of account information of the target account, determining the account risk index corresponding to the target account by using a security evaluation model trained by a neural network, and according to the plurality of article information of the articles published by the target account in the preset time period, determining the article risk index corresponding to the target account;

[0011] According to the account risk index and the article risk index, determining the network position content security index corresponding to the target account;

[0012] Based on the network position content security index, generating an evaluation result.

[0013] Optionally, the security evaluation model trained by the neural network determines the account risk index corresponding to the target account according to the plurality of account information of the target account, comprising:

[0014] Obtaining the account risk weight and the first risk index corresponding to each account information by using the security evaluation model trained by the neural network;

[0015] For each account information, a first product between an account risk weight corresponding to the account information and a preset value is calculated;

[0016] A difference between the first product and a first risk index corresponding to the account information is determined as a second risk index corresponding to the account information;

[0017] Second risk indexes corresponding to each account information are summed up to obtain an account risk index corresponding to the target account.

[0018] Optionally, according to a plurality of article information of an article published by the target account in a preset time period, the security evaluation model trained through the neural network determines a publishing risk index corresponding to the target account, including:

[0019] The security evaluation model trained through the neural network obtains a publishing risk weight and a third risk index corresponding to each article information;

[0020] For each article information, a second product between the publishing risk weight corresponding to the article information and a preset value is calculated;

[0021] A difference between the second product and a third risk index corresponding to the article information is determined as a fourth risk index corresponding to the article information;

[0022] Fourth risk indexes corresponding to each article information are summed up to obtain a publishing risk index corresponding to the target account.

[0023] Optionally, the network position content security index corresponding to the target account is determined according to the account risk index and the publishing risk index, including:

[0024] The network position content security index corresponding to the target account is determined through the following formula:

[0025] PPSI=ARI+f(T MAX -t)*NARI,t≤T MAX

[0026] Wherein, PPSI is the network position content security index, ARI is the account risk index, NARI is the publishing risk index, T MAX is the time from the publishing time to the time when the account heat drops to the daily average, and t is the length of time from the publishing time to the current time.

[0027] Optionally, the evaluation result is generated based on the network position content security index, including:

[0028] In the case that the network position content security index is in a first value range, it is determined that the target account is in a stable period.

[0029] In a case where the network position content security index is in a second numerical range, it is determined that the target account is in an abnormal period;

[0030] In a case where the network position content security index is in a third numerical range, it is determined that the target account is in a pre-warning period;

[0031] The maximum value included in the first numerical range is less than the minimum value included in the second numerical range, and the maximum value included in the second numerical range is less than the minimum value included in the third numerical range.

[0032] Optionally, after the evaluation result is generated based on the network position content security index, the method further comprises:

[0033] In a case where the evaluation result indicates that the target account is in a stable period, if an article published by the target account is monitored, a security evaluation model trained by a neural network is used to re-determine an article risk index corresponding to the target account according to a plurality of article information of the article.

[0034] According to the account risk index corresponding to the target account and the re-determined article risk index, the evaluation result is re-generated.

[0035] Optionally, after the evaluation result is generated based on the network position content security index, the method further comprises:

[0036] In a case where the evaluation result indicates that the target account is in a stable period, the plurality of account information of the target account is updated every interval of a preset time length.

[0037] A security evaluation model trained by a neural network is used to re-determine an account risk index corresponding to the target account according to the updated plurality of account information.

[0038] According to the article risk index corresponding to the target account and the re-determined account risk index, the evaluation result is re-generated.

[0039] Optionally, before the plurality of account information of the target account and the plurality of article information of the article published by the target account in a preset time period are obtained, the method further comprises:

[0040] A plurality of account information and a plurality of article information of a neural network training account are obtained.

[0041] The plurality of account information and the plurality of article information are classified to obtain different types of data sets; the data sets include text data sets, numerical data sets, and category data sets.

[0042] The text data in the text dataset is segmented into words and vectorized to obtain text vectors;

[0043] Determine the risk level of each numerical data point in the numerical dataset, and generate a risk label based on the risk level of each numerical data point;

[0044] The category data in the category dataset is converted into binary vectors using one-hot encoding technology;

[0045] The neural network training data is input into a preset security assessment model, and the security assessment model is iteratively trained using a neural network; the neural network training data includes the text vector, the risk label, and the binary vector;

[0046] When the loss function corresponding to the security assessment model is lower than a preset threshold, a security assessment model trained by the neural network is obtained.

[0047] Furthermore, to achieve the above objectives, this application also provides an account security assessment device, which includes:

[0048] The first acquisition module is used to acquire multiple account information of the target account, as well as multiple article information of articles published by the target account within a preset time period;

[0049] The first determining module is used to determine the account risk index corresponding to the target account based on the security assessment model trained by the neural network, and to determine the posting risk index corresponding to the target account based on the posting information of the target account based on the posting information of the target account in a preset time period.

[0050] The second determining module is used to determine the network content security index corresponding to the target account based on the account risk index and the posting risk index.

[0051] The first generation module is used to generate evaluation results based on the network site content security index.

[0052] To address the aforementioned technical problems, this application also provides a computer device that employs the following technical solution:

[0053] The computer device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the account security assessment method proposed in any of the embodiments of this application.

[0054] To address the aforementioned technical problems, this application also provides a computer-readable storage medium, employing the technical solution described below:

[0055] The computer readable storage medium stores a computer program, and the computer program, when executed by a processor, implements the steps of the account security assessment method according to any one of the embodiments of the present application.

[0056] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0057] The present application provides an account security assessment method, device and equipment based on neural network, the above-mentioned method comprising: obtaining a plurality of account information of a target account, and a plurality of article information of articles published by the target account in a preset time period; determining an account risk index corresponding to the target account according to the plurality of account information of the target account, and determining an article publishing risk index corresponding to the target account according to the plurality of article information of the articles published by the target account in the preset time period, by using a security assessment model trained by a neural network; determining a network position content security index corresponding to the target account according to the account risk index and the article publishing risk index; and generating an evaluation result based on the network position content security index. In the embodiments of the present application, the security assessment model trained by the neural network determines the account risk index based on the account information, determines the article publishing risk index based on the article information, and then determines the network position content security index based on the account risk index and the article publishing risk index, and generates the evaluation result based on the network position content security index. In the account security assessment method provided in the embodiments of the present application, the account information does not need to be analyzed manually, so that the evaluation efficiency is improved; no specific mathematical model and statistical method needs to be selected, and the method is universal; and the security assessment model trained by the neural network is used for data processing, so that the accuracy of the evaluation result is improved. BRIEF DESCRIPTION OF DRAWINGS

[0058] In order to more clearly illustrate the schemes in the present application, the drawings needed in the description of the embodiments of the present application will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0059] Figure 1 is an exemplary system architecture diagram to which the present application can be applied;

[0060] Figure 2 is a flowchart of the account security assessment method based on neural network provided by the embodiments of the present application;

[0061] Figure 3 is an application flowchart of the account security assessment method provided by the embodiments of the present application;

[0062] Figure 4is a schematic diagram of a multi-layer fusion model provided by an embodiment of the present application;

[0063] Figure 5 is a loss diagram on a training set and a test set based on a single-layer linear regression model provided by an embodiment of the present application;

[0064] Figure 6 is a loss diagram on a training set and a test set based on a multi-layer fusion model provided by an embodiment of the present application;

[0065] Figure 7 is a structural schematic diagram of an embodiment of an account security assessment device provided by an embodiment of the present application;

[0066] Figure 8 is a basic structural block diagram of a computer device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0067] The account security assessment method based on a neural network provided by an embodiment of the present application is applied to an account security assessment device. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs; the terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "include" and "have" and any variations thereof in the specification and claims of the present application and the above description of drawings are intended to cover non-exclusive inclusion. The terms "first", "second" and the like in the specification and claims of the present application or the above description of drawings are used to distinguish different objects, not to describe a specific order.

[0068] In this paper, the phrase "embodiment" means that the specific features, structures or characteristics described in conjunction with the embodiment can be included in at least one embodiment of the present application. The phrase appears in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment to other embodiments. Those skilled in the art explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0069] In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the drawings.

[0070] As shown in Figure 1 , the system architecture 100 can include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a communication link medium between the terminal devices 101, 102, 103 and the server 105. The network 104 can include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0071] The user can use the terminal device 101, 102, 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social online platform software, etc.

[0072] The terminal device 101, 102, 103 can be various electronic devices with a display screen and supporting web browsing, including but not limited to smartphones, tablet computers, e-book readers, MP3 (Moving Picture Experts Group Audio Layer III) players, MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, desktop computers, etc.

[0073] The server 105 can be a server providing various services, such as a background server supporting the pages displayed on the terminal device 101, 102, 103.

[0074] It should be noted that the account security assessment method provided by the embodiments of the present application is generally executed by the server / terminal device, and accordingly, the account security assessment device is generally arranged in the server / terminal device.

[0075] It should be understood that Figure 1 The number of terminal devices, networks and servers in

[0076] Please refer to Figure 2 , which shows a flowchart of one embodiment of the account security assessment method based on neural network according to the present application.

[0077] The account security assessment method based on neural network provided by the embodiments of the present application includes the following steps:

[0078] S210, obtaining a plurality of account information of a target account, and a plurality of article information of articles published by the target account in a preset time period.

[0079] It should be noted that the above account information collectively constitutes a multi-dimensional and multi-level evaluation framework, which can more accurately capture and quantify the risk characteristics of the account in different aspects, thereby providing strong technical support and decision basis for account risk management. Optionally, the account risk index includes a primary indicator and a secondary indicator, and the above account information includes the secondary indicator. Please refer to Table I:

[0080] Table I:

[0081]

[0082] It should be noted that the article risk index is the risk index of a specific article in a specific account of a specific network position. The purpose of constructing the article risk index is to comprehensively consider the basic information, sensitive elements and article propagation rate of the newly published article, so as to fully reflect the basic situation, sensitive information risk level and propagation effect of the article in the network.

[0083] Optionally, when an article is published, its risk index will be dynamically calculated based on the given indicator formula. However, in order to more reasonably manage the risk, this project sets a specific time window: if the total risk index of the account does not reach the medium-high risk level within 72 hours after the article is published, it can be considered that no new article is generated during this period, and the risk index of the article will be temporarily zero. This setting helps to reduce the excessive influence of historical article cumulative risk on the current state of the account, making the risk management more focused on the latest dynamics of the account. However, it is worth noting that if the publication of an article causes the total risk index of the account to reach the medium-high risk level, even if the time window of 72 hours is exceeded, the article will still be continuously monitored. In this case, if the account needs to exit the warning period and return to normal state, it needs to rely on the sensitive event index for comprehensive evaluation. Sensitive elements are the core content of evaluating article risk, which is directly related to the compliance and health of article content.

[0084] Optionally, the article risk index includes a primary indicator and a secondary indicator, and the above article information includes the secondary indicator. Please refer to Table II:

[0085] Table II:

[0086]

[0087] Among them, the basic information mainly includes three secondary indicators of title name, content form and emotion field:

[0088] (1) Title: The title is the face of the article and the first element that attracts readers to click. However, overly exaggerated, sensitive words or misleading titles often easily cause readers' dissatisfaction and even adverse effects. Therefore, the title is given a negative weight in the risk assessment to remind the author to pay attention to the compliance and accuracy of the title.

[0089] (2) Content form: The content form is the specific way the article presents to the user, including text, pictures, videos and other forms. Different content forms have different attractions and influences on users.

[0090] (3) Emotional field: Emotion is an important dimension of the information conveyed by the article, which directly affects the reading experience and emotional resonance of the reader.

[0091] The above content defines the account information and article information of the account.

[0092] In this step, the target account's multiple account information and the target account's multiple article information published in a preset time period are obtained. Optionally, the above-mentioned preset time period is 72 hours.

[0093] S220, through the security evaluation model trained by the neural network, determines the account risk index corresponding to the target account according to the multiple account information of the target account, and determines the article risk index corresponding to the target account according to the multiple article information published by the target account in a preset time period.

[0094] It should be noted that the security evaluation model can be divided into four levels, as follows:

[0095] (1) Target layer:

[0096] This layer is the highest layer of the entire evaluation model, and its core goal is to directly output the risk value of the network position account. This risk value is a quantitative reflection of the comprehensive risk assessment of the network position account, which reflects the security status of the account simply and clearly by considering multiple dimensions and characteristics of the account. This layer uses machine learning algorithms such as neural networks or linear regression models to fuse and calculate the data from the lower layers, and finally obtains the risk value of the account.

[0097] (2) Strategy layer:

[0098] In the strategy layer, we consider various indicators of network position accounts from multiple strategic perspectives. Mainly including the indicators of the account itself, the indicators of the published article content, the records of the historical warning library and the relevance of the current sensitive events and the behavior of the account. These four dimensions together constitute the evaluation framework of the strategy layer, providing direction for subsequent detailed analysis.

[0099] (3) Primary indicator layer:

[0100] In the primary indicator layer of network position content security index, multiple key indicators are covered, including login device information, historical situation, account propagation rate, basic attributes, sensitive element analysis, article popularity, and text description. Login device information reflects the usage environment and security of the account; historical situation reveals the past behavior patterns and potential risk points of the account; account propagation rate shows the propagation ability and influence of the account content; basic attributes cover the basic information and settings of the account; sensitive element analysis is used to identify potential risk elements in the account content; article popularity reflects the popularity and social influence of the account published content; and text description is a detailed text analysis of the account published content. These primary indicators together constitute a comprehensive and in-depth assessment of the account, providing a basis for the refinement of secondary indicators.

[0101] (4) Secondary indicator layer:

[0102] The account-level risk index determines the security level of each account under different indicators. Further, the article risk index of the account will further determine the internal security level of each account. In order to effectively evaluate the risk level of the article, the system covers three primary indicators: basic information, article popularity, and sensitive elements, and further subdivides them into multiple secondary indicators to achieve precise quantification of article risk.

[0103] In this step, the security evaluation model trained by the neural network processes multiple account information to obtain the account risk index corresponding to the target account. The account risk index represents the account risk of the target account.

[0104] In this step, the security evaluation model trained by the neural network processes multiple article information published by the target account in a preset time period to obtain the article risk index corresponding to the target account. The article risk index represents the risk of the articles published by the target account.

[0105] Wherein, the account risk index is also referred to as ARI (Account Risk Index) index, and the article risk index is also referred to as NARI (New Article Risk Index) index. The security evaluation model is obtained by neural network training.

[0106] S230, according to the account risk index and the article risk index, determine the network position content security index corresponding to the target account.

[0107] In this step, after obtaining the account risk index and the post risk index, the network position content safety index corresponding to the target account can be determined according to the account risk index and the post risk index. For specific implementation, please refer to subsequent examples.

[0108] The network position content safety index is also referred to as PPSI (Positions and Platform Safety Index).

[0109] In S240, an evaluation result is generated based on the network position content safety index.

[0110] In this step, after obtaining the network position content safety index, an evaluation result can be generated based on the numerical interval in which the network position content safety index is located.

[0111] Optionally, the evaluation result includes three results: the target account is in a stable period, the target account is in an abnormal period, and the target account is in a warning period.

[0112] (1) Stable period: This stage is considered a low-risk period. During this period, the account's activities exhibit stable and compliant characteristics, and no special attention or additional risk management measures are required. Under normal circumstances, the account defaults to this stage and enjoys a relatively relaxed management environment.

[0113] (2) Abnormal period: This marks the account entering a medium-risk period. During this stage, the account behavior exhibits certain abnormalities, and there is a suspected potential risk. Therefore, it is recommended to conduct further manual research and judgment on such accounts, combined with machine learning technology, to conduct in-depth analysis of the account, so as to more accurately identify and respond to possible problems.

[0114] (3) Warning period: The account is considered to be in a high-risk period in this stage. In this critical stage, emergency measures must be taken to prevent potential adverse effects.

[0115] The application provides a method for evaluating the security of an account. The method comprises: obtaining a plurality of account information of a target account and a plurality of article information of articles published by the target account in a preset time period; determining an account risk index corresponding to the target account according to the plurality of account information of the target account by using a security evaluation model trained by a neural network, and determining an article risk index corresponding to the target account according to the plurality of article information of the articles published by the target account in the preset time period; determining a network position content security index corresponding to the target account according to the account risk index and the article risk index; and generating an evaluation result based on the network position content security index. In the embodiment of the application, the security evaluation model trained by the neural network determines the account risk index according to the account information, determines the article risk index according to the article information, and then determines the network position content security index based on the account risk index and the article risk index, and generates the evaluation result based on the network position content security index. In the method for evaluating the security of an account provided in the embodiment of the application, the account information does not need to be analyzed manually, so that the evaluation efficiency is improved; no specific mathematical model and statistical method needs to be selected, so that the method is universal; and the security evaluation model trained by the neural network is used for data processing, so that the accuracy of the evaluation result is improved.

[0116] Optionally, the security evaluation model trained by the neural network determines the account risk index corresponding to the target account according to the plurality of account information of the target account, and comprises:

[0117] The security evaluation model trained by the neural network obtains an account risk weight and a first risk index corresponding to each account information;

[0118] For each account information, a first product between the account risk weight corresponding to the account information and a preset value is calculated;

[0119] A difference between the first product and a first risk index corresponding to the account information is determined as a second risk index corresponding to the account information;

[0120] The second risk indices corresponding to each account information are summed to obtain the account risk index corresponding to the target account.

[0121] Optionally, the second risk index corresponding to the account information can be calculated by the following formula:

[0122] x' k =ω k *z-score(x k )

[0123] wherein x' k represents the second risk index, x k represents the first risk index, and ωk represents the account risk weight, and z represents a preset value.

[0124] Optionally, the account risk index corresponding to the target account can be calculated by the following formula:

[0125]

[0126] wherein, ARI represents the account risk index, x' k represents the second risk index, and k represents the number of account information.

[0127] In the embodiment, the account risk index corresponding to the target account is determined according to the risk situation of each account information, so that the above account risk index can accurately reflect the risk of the target account itself.

[0128] Optionally, the security evaluation model trained by the neural network determines the article risk index corresponding to the target account according to a plurality of article information of articles published by the target account in a preset time period, and the article risk index includes:

[0129] The security evaluation model trained by the neural network obtains the article risk weight and the third risk index corresponding to each article information;

[0130] For each article information, a second product between the article risk weight corresponding to the article information and a preset value is calculated;

[0131] A difference value between the second product and the third risk index corresponding to the article information is determined as a fourth risk index corresponding to the article information;

[0132] The fourth risk index corresponding to each article information is summed to obtain the article risk index corresponding to the target account.

[0133] Optionally, the fourth risk index corresponding to the article information can be calculated by the following formula:

[0134] y' m = ω' m * z-score (y m )

[0135] wherein, y′ m represents the fourth risk index, y m represents the third risk index, ω m ' represents the account risk weight, and z represents a preset value.

[0136] Optionally, the article risk index corresponding to the target account can be calculated by the following formula:

[0137]

[0138] wherein NARI represents a post risk index, y' represents a fourth risk index, and m represents a number of article information. m

[0139] In this embodiment, the post risk index corresponding to the target account is determined according to the risk situation of the article published by the target account, so that the post risk index can accurately reflect the risk of the article published by the target account.

[0140] Optionally, the network position content security index corresponding to the target account is determined based on the account risk index and the post risk index, and the method comprises:

[0141] The network position content security index corresponding to the target account is determined by the following formula:

[0142] PPSI = ARI + f (T MAX -t) * NARI, t≤T MAX

[0143] wherein PPSI represents a network position content security index, ARI represents an account risk index, NARI represents a post risk index, T MAX represents a time from the post time to the time when the account popularity drops to the daily average, and t represents a time length from the post time to the current time.

[0144] In this embodiment, the account risk index can accurately reflect the risk of the target account itself, and the post risk index can accurately reflect the risk of the article published by the target account, so that the network position content security index determined based on the account risk index and the post risk index can accurately represent the risk of the target account.

[0145] Optionally, the evaluation result is generated based on the network position content security index, and the method comprises:

[0146] In a case where the network position content security index is in a first numerical range, it is determined that the target account is in a stable period.

[0147] In a case where the network position content security index is in a second numerical range, it is determined that the target account is in an abnormal period.

[0148] In a case where the network position content security index is in a third numerical range, it is determined that the target account is in a pre-warning period.

[0149] wherein a maximum value included in the first numerical range is less than a minimum value included in the second numerical range, and a maximum value included in the second numerical range is less than a minimum value included in the third numerical range. ​

[0150] In this embodiment, a first numerical range, a second numerical range and a third numerical range are pre-set, wherein the maximum value included in the first numerical range is less than the minimum value included in the second numerical range, and the maximum value included in the second numerical range is less than the minimum value included in the third numerical range.

[0151] In the case where the network site content security index is in the first numerical range, the activity of the account shows stable and compliant characteristics, and no special attention or additional risk management measures need to be taken, it is determined that the target account is in a stable period.

[0152] In the case where the network site content security index is in the second numerical range, the behavior of the account shows a certain abnormality, and there is a suspected potential risk, it is determined that the target account is in an abnormal period.

[0153] In the case where the network site content security index is in the third numerical range, the account is considered to be in a high-risk period at this stage, it is determined that the target account is in a warning period.

[0154] Optionally, after the evaluation result is generated based on the network site content security index, the method further comprises:

[0155] In the case where the evaluation result represents that the target account is in a stable period, if an article published by the target account is monitored, a security evaluation model trained by a neural network is used to re-determine an article risk index corresponding to the target account according to a plurality of article information of the article.

[0156] According to the account risk index corresponding to the target account and the re-determined article risk index, the evaluation result is re-generated.

[0157] In this embodiment, in the case where the target account is in a stable period, the target account is monitored in real time, if an article published by the target account is monitored, a security evaluation model trained by a neural network is used to re-determine an article risk index corresponding to the target account according to a plurality of article information of the article.

[0158] Further, according to the account risk index corresponding to the target account and the re-determined article risk index, the evaluation result is re-generated, and the specific implementation manner can refer to the above-mentioned embodiments.

[0159] Optionally, after the evaluation result is generated based on the network site content security index, the method further comprises:

[0160] In the case where the evaluation result represents that the target account is in a stable period, the plurality of account information of the target account is updated every interval of a preset time length.

[0161] The security evaluation model trained by the neural network determines the account risk index corresponding to the target account according to the updated multiple account information.

[0162] According to the article risk index corresponding to the target account and the re-determined account risk index, the evaluation result is re-generated.

[0163] In the embodiment, when the target account is in the stable period, the multiple account information of the target account is updated every interval preset time, and the security evaluation model trained by the neural network determines the account risk index corresponding to the target account according to the updated multiple account information. Further, according to the article risk index corresponding to the target account and the re-determined account risk index, the evaluation result is re-generated. For specific implementation, please refer to the above embodiment.

[0164] For the purpose of illustrating the above embodiments, please refer to Figure 3 , in which Figure 3 In the example shown, in the stable period of the target account, if the target account does not post articles, the account information of the target account is updated every interval preset time, and the account risk index is recalculated. Based on the account risk index, it is determined whether the target account is at risk. If there is no risk, it is determined that the target account is in the stable period; if there is a risk, it is determined that the target account is in the abnormal period, and the account risk of the target account is evaluated again every interval preset time.

[0165] In the stable period of the target account, if the target account posts articles, the article risk index of the newly published article of the target account is calculated, and according to the above article risk index and the account risk index of the target account, it is determined whether the target account is at risk. If not, it is determined that the target account is in the stable period; if so, it is determined that the target account is in the warning period.

[0166] When the target account is in the warning period, if a sensitive event occurs, the sensitive event index of the sensitive event is determined, and the state of the target account is determined according to the sensitive event index.

[0167] The sensitive event is a secondary index included in the article risk index.

[0168] Optionally, different sensitive events can correspond to different sensitive event indexes.

[0169] Optionally, before the multiple account information of the target account and the multiple article information published by the target account in the preset time period are obtained, the method further comprises:

[0170] Obtain the multiple account information and the multiple article information of the neural network training account;

[0171] The plurality of account information and the plurality of article information are classified to obtain different types of data sets; the data sets include text data sets, numerical data sets, and category data sets;

[0172] The text data in the text data set is segmented and vectorized to obtain a text vector;

[0173] The risk level of each numerical data in the numerical data set is determined, and a risk label is generated based on the risk level of each numerical data;

[0174] The category data in the category data set is converted into a binary vector through a one-hot encoding technique;

[0175] The neural network training data is input into a preset security evaluation model to iteratively train the security evaluation model; the neural network training data includes the text vector, the risk label, and the binary vector;

[0176] If the loss function corresponding to the security evaluation model is lower than a preset threshold, a security evaluation model with completed neural network training is obtained.

[0177] In this embodiment, the plurality of account information and the plurality of article information of the neural network training account are obtained, and the plurality of account information and the plurality of article information are classified to obtain different types of data sets. Optionally, the above data sets include text data sets, numerical data sets, and category data sets. The above different types of data are subjected to corresponding data preprocessing, specifically:

[0178] (1) Text data: In the preprocessing stage of text data processing, irrelevant characters are removed to clean the data, and word segmentation is performed to facilitate subsequent text analysis and model training. Specifically, the jieba tool based on dynamic programming algorithm is used for Chinese word segmentation, which can effectively segment the title text into meaningful lexical units. In order to maintain the consistency of the data, the maximum number of words in each title is set to 10. For titles with less than 10 words, a zero padding method is used to ensure that each title has the same dimension in subsequent processing. After completing the word segmentation, each word is further vectorized using Tencent's Word2Vec database. Word2Vec is an advanced word embedding technology that can map words to vector representations in high-dimensional space, capturing the semantic relationships between words. Optionally, each word is projected into a 100-dimensional vector. Therefore, each title composed of 10 words is finally represented as a 1000-dimensional vector. This vector representation not only preserves the semantic information of the words, but also provides convenience for subsequent text analysis and model neural network training.

[0179] (2) Numerical data: Numerical data is a decisive factor for content security. In the risk assessment of data labels, the total number of rejections is mainly used. Specifically, when the total number of rejections of a certain content is greater than 0, it is assigned a risk label of 1, indicating that the content is at risk. Conversely, when the total number of rejections is 0, the risk label is 0, indicating that the content is highly secure and does not present obvious risks. This risk assessment mechanism helps quickly and accurately identify potential risk content during content review and management, so that appropriate measures can be taken to ensure the safety of platform content. In addition, a certain threshold is set for the total number of suspected cases. When the total number of suspected cases exceeds this specific threshold, a risk warning mechanism is triggered to further review and handle potential risk content.

[0180] (3) Category data: In the category data processing stage, a one-hot encoding technique is used to convert categorical data into numerical form. Hot encoding is a common method for processing categorical data, which converts each categorical variable into a binary vector. In this vector, the position representing the current category is set to 1, while all other positions are set to 0. This method effectively processes non-numerical categorical data and converts it into a numerical form that the model can understand. After completing the numerical conversion of the data, all data is further standardized using the Z-Score standardization method. Z-Score standardization is a common data preprocessing technique that converts data to a standard distribution form with a mean of 0 and a standard deviation of 1 by subtracting the mean and dividing by the standard deviation. This step is crucial for eliminating differences between data dimensions, improving model convergence speed, and enhancing model generalization ability.

[0181] Further, the neural network training data is input into the preset security evaluation model for iterative training of the security evaluation model.

[0182] It should be understood that the security evaluation model includes a single-layer linear regression model, i.e., the security evaluation model is constructed based on a single-layer linear regression model.

[0183] First, a single-layer linear regression model in a neural network is used to estimate the weight factor in different influencing factors, which can be represented as:

[0184] Y = Wx + b,

[0185] where W represents the weight matrix of the influence of different influencing factors on content security, bwhere b is the bias coefficient, x is the model input data, and Y represents the output class probability. In this model, the backpropagation algorithm is used as the parameter optimization method. Backpropagation is an efficient gradient descent technique that calculates the gradient of the loss function with respect to each parameter and updates the parameter values in the opposite direction of the gradient, thereby minimizing the loss function and optimizing the model.

[0186] In this model, the input data x encompasses pre-processed text data, numerical data, and categorical data, which collectively form the feature set of the model. The output variable Y is determined based on the total number of rejections in the numerical data. Specifically, when the total number of rejections is greater than 0, the risk is labeled as 1, indicating the presence of risk; when the total number of rejections is 0, we label the risk as 0, indicating no risk. This processing method enables the model to predict the corresponding account risk status based on the input feature data, providing strong support for subsequent platform risk management and decision-making. After optimization by the backpropagation algorithm, a weight matrix is obtained, where each element represents the importance of the corresponding content security impact parameter. This result not only reveals the influence degree of different factors on the target variable, but also provides a quantitative basis for the subsequent formulation of different network site content security. In addition, the optimization process involves iterative calculation of the loss function and fine-tuning of the model parameters based on gradient information to ensure the accuracy and stability of the model prediction.

[0187] It should be understood that the security assessment model includes a multilayer linear regression model, i.e., the security assessment model is constructed based on the multilayer linear regression model.

[0188] The multilayer linear regression model includes 7 pre-connected fully connected layers, which correspond to the article title, main unit, content type, usage frequency, location (inland / overseas), whether there is a content review mechanism, and other potential influencing factors in the multi-dimensional influence factors. The core function of this model is that it can effectively reduce the multi-dimensional vector features to one-dimensional space, and then calculate the weights through a unified fully connected layer. The weights derived by this model can be uniquely mapped to each feature in the multi-dimensional influence factors, greatly enhancing the interpretability of the model. This feature makes the model more practical in actual application, such as Figure 4 As shown, it clearly shows how the model effectively maps multi-dimensional features to one-dimensional weight space, improving the interpretability and practicality of the model.

[0189] In addition, in the design of the full connection layer of the multi-layer linear regression model, the ReLU activation function is used to enhance the non-linear expression ability of the model, and the Dropout technology is used to reduce the risk of overfitting, ensuring the generalization ability of the model on complex data sets. In addition, in order to further improve the stability and convergence of the model during the neural network training process, a batch normalization layer is added after each full connection layer, which helps to normalize the input distribution and reduce internal covariate shift.

[0190] Further, in the case that the loss function corresponding to the security evaluation model is lower than the preset threshold, the security evaluation model trained by the neural network is obtained.

[0191] To prove the accuracy of the output result of the security evaluation model, please refer to Figure 5 and Figure 6 . 180,000 data points are randomly selected from the complete data set as the neural network training set, and are verified on an independent test data set. The verification result of the single-layer linear regression model is shown in Figure 5 . During the neural network training process, as the sample size increases, the loss on the neural network training set and the test set gradually decreases, indicating that the accuracy of the security evaluation model has been improved.

[0192] The verification result of the multi-layer linear regression model is shown in Figure 6 . Since the optimization algorithm and the regularization technique are used, the generalization ability based on the multi-layer linear regression model is effectively ensured.

[0193] Please refer to Figure 7 , the account security evaluation device 700 provided by the embodiment of the application comprises:

[0194] The first acquisition module 710 is configured to acquire a plurality of account information of a target account and a plurality of article information of articles published by the target account in a preset time period;

[0195] The first determination module 720 is configured to determine, by using the security evaluation model trained by the neural network, an account risk index corresponding to the target account according to the plurality of account information of the target account, and determine an article publishing risk index corresponding to the target account according to the plurality of article information of the articles published by the target account in the preset time period;

[0196] The second determination module 730 is configured to determine a network position content security index corresponding to the target account according to the account risk index and the article publishing risk index.

[0197] The first generation module 740 is configured to generate an evaluation result based on the network position content security index.

[0198] Optionally, the first determining module 720 is specifically used for:

[0199] The security evaluation model trained by the neural network obtains an account risk weight corresponding to each account information and a first risk index;

[0200] For each account information, a first product between the account risk weight corresponding to the account information and a preset value is calculated;

[0201] A difference between the first product and a first risk index corresponding to the account information is determined as a second risk index corresponding to the account information;

[0202] Second risk indexes corresponding to each account information are summed to obtain an account risk index corresponding to the target account.

[0203] Optionally, the first determining module 720 is specifically used for:

[0204] The security evaluation model trained by the neural network obtains an article risk weight corresponding to each article information and a third risk index;

[0205] For each article information, a second product between the article risk weight corresponding to the article information and a preset value is calculated;

[0206] A difference between the second product and a third risk index corresponding to the article information is determined as a fourth risk index corresponding to the article information;

[0207] Fourth risk indexes corresponding to each article information are summed to obtain an article risk index corresponding to the target account.

[0208] Optionally, the second determining module 730 is specifically used for:

[0209] The network position content security index corresponding to the target account is determined by the following formula:

[0210] PPSI = ARI + f(T MAX -t) * NARI, t ≤ T MAX

[0211] Wherein, PPSI is the network position content security index, ARI is the account risk index, NARI is the article risk index, T MAX is the time from the article publishing time to the time when the account popularity drops to the daily average, and t is the time length from the article publishing time to the current time.

[0212] Optionally, the first generating module 740 is specifically used for:

[0213] In a case where the network position content security index is in a first numerical range, it is determined that the target account is in a stable period;

[0214] In a case where the network position content security index is in a second numerical range, it is determined that the target account is in an abnormal period;

[0215] In a case where the network position content security index is in a third numerical range, it is determined that the target account is in a pre-warning period.

[0216] The first numerical range includes a maximum value that is less than a minimum value included in the second numerical range, and the second numerical range includes a maximum value that is less than a minimum value included in the third numerical range.

[0217] Optionally, the account security evaluation device 700 further includes:

[0218] The third determination module is configured to, in a case where the evaluation result indicates that the target account is in a stable period, if an article published by the target account is monitored, determine, by the security evaluation model trained by the neural network, an article risk index corresponding to the target account according to a plurality of article information of the article.

[0219] The second generation module is configured to regenerate the evaluation result according to the account risk index corresponding to the target account and the article risk index determined again.

[0220] Optionally, the account security evaluation device 700 further includes:

[0221] The update module is configured to, in a case where the evaluation result indicates that the target account is in a stable period, update a plurality of account information of the target account every interval of a preset time length.

[0222] The fourth determination module is configured to determine, by the security evaluation model trained by the neural network, an account risk index corresponding to the target account according to the plurality of updated account information.

[0223] The third generation module is configured to regenerate the evaluation result according to the article risk index corresponding to the target account and the account risk index determined again.

[0224] Optionally, the account security evaluation device 700 further includes:

[0225] The second acquisition module is configured to acquire a plurality of account information and a plurality of article information of a neural network training account.

[0226] The classification module is configured to classify the plurality of account information and the plurality of article information to obtain different types of data sets; the data sets include text data sets, numerical data sets and category data sets.

[0227] The processing module is configured to perform word segmentation and vectorization processing on the text data in the text data sets to obtain text vectors.

[0228] The fifth determination module is configured to determine risk levels of the numerical data in the numerical data sets, and generate risk labels based on the risk levels of the numerical data.

[0229] The conversion module is configured to convert the category data in the category data sets into binary vectors by using a one-hot encoding technique.

[0230] The neural network training module is configured to input neural network training data into a preset security evaluation model, and perform iterative neural network training on the security evaluation model; the neural network training data include the text vectors, the risk labels and the binary vectors.

[0231] The sixth determination module is configured to obtain a trained security evaluation model when a loss function corresponding to the security evaluation model is lower than a preset threshold.

[0232] To solve the above technical problems, the embodiments of the present application further provide a computer device. For details, please refer to Figure 8 , Figure 8 The basic structure block diagram of the computer device of the present embodiment is shown in the figure.

[0233] The computer device 8 includes a memory 81, a processor 82 and a network interface 83 which are connected to each other through a system bus. It should be pointed out that only the computer device 8 with components 81-83 is shown in the figure, but it should be understood that all the shown components are not required to be implemented, and more or fewer components can be alternatively implemented. Among them, those skilled in the art can understand that the computer device herein is a device capable of automatically performing numerical calculation and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0234] The computer device can be a desktop computer, a notebook computer, a palm computer, a cloud server, or the like. The computer device can interact with a user through a keyboard, a mouse, a remote controller, a touchpad, a voice control device, or the like.

[0235] The memory 81 can include at least one type of readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory, or the like), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, or the like. In some embodiments, the memory 81 can be an internal storage unit of the computer device 8, such as a hard disk or a memory of the computer device 8. In other embodiments, the memory 81 can also be an external storage device of the computer device 8, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, or the like. Of course, the memory 81 can include both an internal storage unit and an external storage device of the computer device 8. In this embodiment, the memory 81 is generally used to store an operating system and various application software installed in the computer device 8, such as program codes of the account security assessment method, or the like. In addition, the memory 81 can also be used to temporarily store various data that have been output or will be output.

[0236] The processor 82 can be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip in some embodiments. The processor 82 is generally used to control the overall operation of the computer device 8. In this embodiment, the processor 82 is used to run program codes or process data stored in the memory 81, such as running program codes of the account security assessment method.

[0237] The network interface 83 can include a wireless network interface or a wired network interface, and is generally used to establish a communication connection between the computer device 8 and other electronic devices.

[0238] The present application also provides another embodiment, i.e., a computer readable storage medium storing the account security assessment program. The account security assessment program can be executed by at least one processor to enable the at least one processor to perform the steps of the account security assessment method as described above.

[0239] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned example method can be realized by means of software and necessary general hardware online platform, of course, it can also be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes a plurality of instructions for making a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) execute the method described in each embodiment of the present application.

[0240] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld devices or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in a distributed computing environment, in which tasks are performed by remote processing devices connected by a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0241] Obviously, the above-described embodiments are only a part of the embodiments of the present application, and are not all the embodiments. The preferred embodiments of the present application are given in the drawings, but do not limit the patent scope of the present application. The present application can be realized in many different forms, and conversely, the purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent replacements to some technical features. Any equivalent structure made by using the contents of the specification and drawings, directly or indirectly applied to other related technical fields, is also within the scope of the patent protection of the present application.

Claims

1. A neural network-based account security assessment method, characterized in that, The method comprises: obtaining a plurality of account information of a target account and a plurality of article information of articles published by the target account in a preset time period; determining, by a security evaluation model trained by a neural network, an account risk index corresponding to the target account according to the plurality of account information of the target account, and determining a publishing risk index corresponding to the target account according to the plurality of article information of the articles published by the target account in the preset time period; determining a network position content security index corresponding to the target account according to the account risk index and the publishing risk index; generating an evaluation result based on the network position content security index; wherein the security evaluation model trained by the neural network determines the account risk index corresponding to the target account according to the plurality of account information of the target account, comprising: obtaining, by the security evaluation model trained by the neural network, an account risk weight and a first risk index corresponding to each account information; for each account information, calculating a first product between the account risk weight corresponding to the account information and a preset value; determining a difference between the first product and a first risk index corresponding to the account information as a second risk index corresponding to the account information; summing up the second risk index corresponding to each account information to obtain the account risk index corresponding to the target account.

2. The method of claim 1, wherein, The security evaluation model trained by the neural network determines the publishing risk index corresponding to the target account according to the plurality of article information of the articles published by the target account in the preset time period, comprising: obtaining, by the security evaluation model trained by the neural network, a publishing risk weight and a third risk index corresponding to each article information; for each article information, calculating a second product between the publishing risk weight corresponding to the article information and a preset value; determining a difference between the second product and a third risk index corresponding to the article information as a fourth risk index corresponding to the article information; summing up the fourth risk index corresponding to each article information to obtain the publishing risk index corresponding to the target account.

3. The method of claim 1, wherein, The security evaluation model trained by the neural network determines the publishing risk index corresponding to the target account according to the plurality of article information of the articles published by the target account in the preset time period, comprising: determining the network position content security index corresponding to the target account by the following formula: Wherein, PPSI is the network site content security index, ARI is the account risk index, NARI is the post risk index, is the time from the post time to the time when the account heat drops to the daily average, t is the time length from the post time to the current time.

4. The method of claim 1, wherein, The security evaluation model trained by the neural network determines the publishing risk index corresponding to the target account according to the plurality of article information of the articles published by the target account in the preset time period, comprising: in the case that the network position content security index is in a first value range, determining that the target account is in a stable period; in the case that the network position content security index is in a second value range, determining that the target account is in an abnormal period; in the case that the network position content security index is in a third value range, determining that the target account is in a warning period; wherein the maximum value included in the first value range is less than the minimum value included in the second value range, and the maximum value included in the second value range is less than the minimum value included in the third value range.

5. The method of claim 1, wherein, After the evaluation result is generated based on the network position content security index, the method further comprises: In a case where the evaluation result indicates that the target account is in a stable period, if it is monitored that the target account posts an article, a security evaluation model trained by a neural network is used to re-determine a post risk index corresponding to the target account according to a plurality of article information of the article. The evaluation result is re-generated according to the account risk index corresponding to the target account and the re-determined post risk index.

6. The method of claim 1, wherein, After the evaluation result is generated based on the network position content security index, the method further comprises: In a case where the evaluation result indicates that the target account is in a stable period, the plurality of account information of the target account is updated every interval of a preset time length. The security evaluation model trained by the neural network is used to re-determine an account risk index corresponding to the target account according to the plurality of updated account information. The evaluation result is re-generated according to the post risk index corresponding to the target account and the re-determined account risk index.

7. The method of claim 1, wherein, Before the plurality of account information of the target account and the plurality of article information of the article posted by the target account in a preset time period are obtained, the method further comprises: Obtain a plurality of account information and a plurality of article information of a neural network training account. Classify the plurality of account information and the plurality of article information to obtain different types of data sets; the data sets include a text data set, a numerical data set, and a category data set. Carry out word segmentation and vectorization processing on the text data in the text data set to obtain a text vector. Determine the risk level of each numerical data in the numerical data set, and generate a risk label based on the risk level of each numerical data. Convert the category data in the category data set into a binary vector by using a one-hot encoding technology. Input neural network training data into a preset security evaluation model to iteratively train the security evaluation model; the neural network training data includes the text vector, the risk label, and the binary vector. In a case where a loss function corresponding to the security evaluation model is lower than a preset threshold, a security evaluation model trained by a neural network is obtained.

8. An account security assessment device, characterized in that, Comprise: The first acquisition module is configured to obtain a plurality of account information of a target account and a plurality of article information of an article posted by the target account in a preset time period. The first determination module is configured to determine, by a security evaluation model trained by a neural network, an account risk index corresponding to the target account according to the plurality of account information of the target account, and determine a post risk index corresponding to the target account according to the plurality of article information of the article posted by the target account in the preset time period. The second determination module is configured to determine a network position content security index corresponding to the target account according to the account risk index and the post risk index. The first generation module is configured to generate an evaluation result based on the network position content security index. The first determination module is specifically configured to: Obtain an account risk weight and a first risk index corresponding to each account information by the security evaluation model trained by the neural network. For each account information, a first product between a risk weight of an account corresponding to the account information and a preset value is calculated; A difference between the first product and a first risk index corresponding to the account information is determined as a second risk index corresponding to the account information; Second risk indices corresponding to each account information are summed up to obtain an account risk index corresponding to the target account.

9. A computer device, comprising: A computer readable storage medium storing a computer program, wherein the computer program is executed by a processor to implement the steps of the account security evaluation method according to any one of claims 1 to 7.

Images