An intelligent method for revealing the risks of Internet-related crimes
By adopting a big data analysis method based on rules engine in cybercriminal activity detection, the problems of low efficiency and inaccurate results in the existing technology are solved, and rapid and accurate risk disclosure and automated processing are achieved. Improve the investigation efficiency and data processing effect of staff.
Patent Information
- Application Number
- CN202111308154.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-11-05
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2041-11-05
AI Technical Summary
The existing technology lacks fast, accurate and automated analysis methods in the detection of cybercrime activities, resulting in low data processing efficiency, inaccurate results, and inability to effectively respond to changes in business needs.
Using the big data analysis method based on the rules engine, by constructing sensitive thesaurus, explosive formula library and detection rules, combined with text analysis technology, network activity data is monitored, classified and integrated, and the risk index of network accounts is output to achieve automation. and flexible risk revelation.
It realizes rapid and accurate detection and intuitive display of risk warnings for cybercrime activities, reduces the workload of manual verification, improves efficiency and effectiveness, and can flexibly respond to changes in business needs and reduces system maintenance costs.
Smart Images

Figure CN114020991B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of the public security industry, and more specifically to a method for intelligently revealing risks of cybercrimes, which is used for pre-detection and discovery of cybercrime activities in the work of the public security industry. Background Art
[0002] In today's Internet+ and big data era, people's lives are becoming increasingly networked and digitized, and criminals are gradually transferring traditional criminal activities to the Internet. Due to the characteristics of fast mobility and rapid changes in online activities and the opacity of network identities, this poses greater challenges to the staff's prevention and crackdown on crime.
[0003] At present, although staff can obtain online activity data in their daily network activity investigation work, due to the large amount of data and the large amount of manual work involved in data classification, analysis and identification, there is a lack of scientific, fast and effective intelligence analysis methods for network activity data, which affects both efficiency and effectiveness.
[0004] Currently, some large Internet companies on the market have some software related to monitoring and detection methods for network activities, but they are all based on general sensitive word matching technology to do some general comparison processing, and the comparison results are directly pushed to relevant units as clues. The clues are not organized, and have not been transformed into account-centric and intuitive and easy-to-understand risk indexes from the perspective of risk management; and there is no combination with related businesses, no targeted denoising and sensitive word replacement mechanism, resulting in problems such as inaccurate clues, large number of pushed clues, and a large workload of manual intervention; at the same time, there is no risk index of the account related to the network behavior from the perspective of risk control and points, and it is impossible to intuitively tell the staff which type of network activities they should pay more attention to; lack of custom configurable rule engine technology, unable to respond quickly and cost-effectively to changes in subsequent user needs.
[0005] Faced with a large amount of network activity data, how to use fast, efficient and automatic technical means to quickly and accurately identify which network behaviors are involved in cybercrime activities? Can some specific rules be formulated in combination with related businesses to filter out some noise? Can this be achieved through an automated and flexibly configurable rule-based engine technology? What are their risk indexes? Which activities and related accounts should the staff focus on? These are all issues that the staff are concerned about and that need to be resolved urgently. They are also a powerful means to reduce the staff's manual verification and analysis of each network activity data, and a scientific way to effectively prevent Internet-related criminal activities in advance. Summary of the invention
[0006] The technical problem to be solved by the present invention is to provide a method for intelligently revealing the risks of Internet-related crimes, so as to solve the problems in the background technology, so as to quickly respond to changes in business needs, and to make the prompts of Internet-related crime risks more intuitive. It can also set relevant scoring and exclusion methods through a flexible rule engine to make the detection operation more automatic and accurate.
[0007] In order to solve the above technical problems, the technical solutions adopted by the present invention are as follows.
[0008] A method for intelligently revealing the risks of Internet-related crimes monitors and classifies the text content of relevant network activity data, combines it with a defined basic library, uses a rule engine, big data comparison, and text analysis technology to extract, analyze, and integrate the relevant network activity data, outputs the relevant scores of the series of network activities, and accumulates the points to the relevant network accounts. After risk level conversion, the final risk level of the account is obtained.
[0009] Further optimize the technical solution, specifically including the following steps:
[0010] S1. Conduct basic construction, including building a vocabulary, building an explosive formula library, and formulating detection rules;
[0011] S2. Define the "integrated Internet-related crime risk disclosure process" model, and start the "integrated Internet-related crime risk disclosure process" model to conduct risk measurement and processing on Internet-related activity data.
[0012] To further optimize the technical solution, in step S1, constructing the vocabulary includes the following steps:
[0013] S101. Sort out and establish a database of sensitive words related to explosive online activities, and classify sensitive words;
[0014] S102. Sort out and collect civilian alternative explosive materials, establish a civilian alternative word library, and set a one-to-one correspondence with the sensitive word library of online activities related to explosives;
[0015] S103. Sort out activities that are related to explosive-making activities but not suspected of crimes, establish a corresponding exclusion word library, and set an exclusion method for each exclusion word.
[0016] To further optimize the technical solution, in step S1, constructing an explosive formula library includes the following steps: according to the preparation requirements of each type of explosive, sorting out the raw materials required for the explosive making formula, establishing a relevant formula table for each type of explosive, and establishing an explosive formula library.
[0017] To further optimize the technical solution, in step S1, formulating the detection rules includes the following steps:
[0018] S111. Formulate scoring rules. According to different types of characteristics, the scoring rules are divided into two categories: "scoring by number of hits" and "scoring by number / weight of hits";
[0019] S112. Formulate formula rules. According to the raw material requirements for explosives, after establishing a relevant formula library, formulate relevant weighted calculation rules for each formula.
[0020] To further optimize the technical solution, step S2 includes the following specific steps:
[0021] S21. Internet data access: obtaining and aggregating Internet data;
[0022] S22, data preprocessing;
[0023] S23, data content extraction;
[0024] S24, loading basic library: loading the basic library information defined in step S21;
[0025] S25, score superposition calculation and weighted processing;
[0026] Score superposition operation: the content extracted from the data content in step S23 is compared and matched according to the category to which the sensitive words belong, and the hit content is scored according to the category to which the sensitive words belong;
[0027] Weighted processing: For online behaviors involving a certain account, if multiple data types are involved and sensitive words or civilian substitute words under the data type are hit, the total score of the account needs to be weighted averaged and then additional points are added to the total score;
[0028] S26. Obtain and output the risk index of the account involved in Internet activities.
[0029] To further optimize the technical solution, step S22 includes the following steps:
[0030] Identify the types of Internet-related data, perform standardized data processing and storage on this type of data, and match the data with corresponding rules.
[0031] To further optimize the technical solution, in step S25, the score superposition operation includes the following steps:
[0032] S251. Online speech data must be processed according to the “hit count”;
[0033] S252: If the same sensitive word is hit again, the score will continue to increase;
[0034] S253. If other sensitive words in the same category are hit, points will continue to be added;
[0035] S254. If multiple hit sensitive words in the same category match the formula library, additional scoring is performed according to the corresponding formula rules;
[0036] S255. When a sensitive word in a different category is hit, in addition to scoring the sensitive word, an additional scoring process is performed;
[0037] S256, when the same data also hits the exclusion word, no score or score reduction is performed according to the type of the exclusion word hit;
[0038] S257, finally obtaining the total score of the data, and adding the score to the relevant speech account;
[0039] S258. Repeat steps S251 to S257 until all data are traversed.
[0040] Due to the adoption of the above technical scheme, the technical progress achieved by the present invention is as follows.
[0041] The present invention is a method for calculating, detecting and analyzing the risks of Internet-related crimes based on Internet activity data. It is a new method based on the integration of multiple technologies such as big data computing, self-developed rule engines and text analysis. It can not only quickly respond to changes in business needs, but also make the prompts of Internet-related crime risks more intuitive. It can also set relevant scoring and exclusion methods through a flexible rule engine to make detection operations more automatic and accurate.
[0042] The present invention can detect text data of online activities such as online speech, communication information, transaction shopping, logistics delivery, etc., obtain the risk index of related accounts through comprehensive calculation, and can quickly and objectively analyze the target account, thereby alleviating the workload of staff who need to manually check network accounts and their activity behaviors one by one, reducing the difficulty of manual analysis and judgment, and allowing staff to quickly predict the account risks involved in the behavior.
[0043] 1. The present invention is based on a custom and configurable rule engine, which fully considers relevant business scenarios, designs integral calculation rules for different scenarios, and more deeply integrates into the business of explosion-related and gun trafficking, and designs different explosive formula rules, materials and accessories combination rules. The rule engine can be adjusted at any time according to changes in future staff needs, which can quickly meet new needs and reduce the development and maintenance costs of related systems.
[0044] 2. The present invention adopts the concept of account risk management and control, and adopts a point calculation mechanism for the detection of Internet-related activities. According to various rules set by the rule engine, combined with weighted processing, points are calculated centered on relevant network accounts. The result is a comprehensive score, which is converted into a corresponding risk level. Compared with the past, staff could only receive scattered clues of suspicious Internet-related criminal activities one by one. The present invention allows staff to identify high-risk and high-score accounts at a glance, and can check related activity data, eliminating the work of manual verification and manual analysis by staff.
[0045] 3. The present invention combines the business of related departments, establishes a related vocabulary, sorts out exclusion words according to multiple dimensions such as different occupations, units, and places, establishes an exclusion mechanism, reduces noise, and reduces the push of useless information to staff.
[0046] 4. The entire analysis method of the present invention adopts an automated processing mechanism. As long as there is continuous access to network activity data, various types of network-related activity data can be automatically analyzed, and high-risk network-related criminal activities and related network accounts can be discovered. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] Figure 1 It is a conceptual block diagram of the risk detection model system of the present invention;
[0048] Figure 2 It is a flow chart of the present invention. DETAILED DESCRIPTION
[0049] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0050] An intelligent method for revealing the risks of Internet-related crimes, combined with Figure 1 to Figure 2 As shown, by monitoring and classifying the text content of relevant network activity data (transactions, speeches, posts, etc.), combining with the defined basic library, using rule engines, big data comparison, text analysis and other technologies, and extracting and analyzing the relevant network activity data through weighted analysis and other methods, and performing integral operations, the relevant scores of the series of network activities are output, and the points are added to the relevant network accounts. After the risk level conversion, the final risk level of the account is obtained, so that staff can promptly discover medium and high-risk network activities and guide the direction of investigation.
[0051] The present invention is guided by the actual business needs of users, takes the prevention and analysis of Internet-related criminal activities as the entry point, takes the detection of explosive-related online activities as an example, uses big data, text analysis, and rule engines as basic technologies, and aims to intelligently, quickly and accurately reveal the risks of Internet-related criminal activities. A set of risk analysis solutions is formed by developing a customized rule engine (integral calculation rules, explosive-related combination formula rules) combined with an automated processing method.
[0052] The method adopted by the present invention can support the risk detection business of network-related explosives and gun trafficking criminal activities. The present invention will use online activities involving explosives as an example to illustrate the processing scheme of the present invention.
[0053] A method for intelligently revealing the risks of Internet-related crimes, comprising the following steps:
[0054] S1: Conduct basic construction, including building a vocabulary, building an explosive formula library, and formulating detection rules. Basic construction is the basis for the subsequent step S2.
[0055] S10. Constructing a word library, including establishing a sensitive word library and classifying sensitive words, establishing a civilian substitute word library and setting a one-to-one correspondence with sensitive words, establishing an exclusion word library and setting exclusion word types, and setting exclusion scoring methods.
[0056] Building a vocabulary includes the following steps:
[0057] S101. Sort out and establish a database of sensitive words related to online explosive activities, and classify these sensitive words according to business requirements and explosive-making activities, which may include two classification modes: large classification and detailed classification.
[0058] S102. According to the needs of explosive-making activities, collect and sort out civilian alternative explosive-making materials, establish a vocabulary of civilian alternatives, and set them in one-to-one correspondence with the sensitive vocabulary of online activities related to explosives.
[0059] S103. Based on the exceptions to explosive-making activities, sort out activities related to explosive-making activities but not suspected of crimes from the perspectives of occupations and units, establish a corresponding exclusion word library, and set exclusion methods for each exclusion word. Exclusion methods include no-score exclusion and minus-score exclusion.
[0060] S11, constructing an explosive formula library, including the following steps: according to the preparation requirements of each type of explosive, sorting out the raw materials required for the explosive formula, establishing a relevant formula table for each type of explosive, and establishing an explosive formula library.
[0061] S12. Formulating detection rules is to conduct content-based analysis of text data generated by online activities. The premise for finding out whether it involves explosives is to formulate corresponding rules.
[0062] Formulate detection rules, including scoring rules and formula combination rules, which specifically include the following steps:
[0063] S121. Formulate scoring rules. The types of online explosive activities are mainly divided into online speech, communication information, transaction shopping, and logistics delivery. According to the characteristics of different types, the scoring rules are defined into two categories: "scoring by number of hits" and "scoring by number / weight of hits".
[0064] The rules for scoring by number of hits are as follows: set the scoring for the first hit of sensitive words, the scoring for the second hit of sensitive words, and the weighted processing scoring for the combined hits of different types of sensitive words according to the classification of sensitive words, and set the scoring rules for exclusion words.
[0065] The rules for scoring by hit quantity / weight are: set the score for hitting sensitive words according to the sensitive word classification and according to different quantity / weight intervals (for example, if a sensitive word is hit in a certain category, 10 points will be given if the number involved is 3-20, and 15 points will be given if the number involved is 20-100), as well as weighted processing and scoring of combined hits of different types of sensitive words, and set scoring rules for excluding words.
[0066] S122. Formulate formula rules. According to the raw material requirements for explosives, after establishing a relevant formula library, formulate relevant weighted calculation rules for each formula so as to be used for weighted scoring calculation when the combination hits different types of sensitive words.
[0067] S2. Define the "integrated Internet-related crime risk disclosure process" model and start the "integrated Internet-related crime risk disclosure process" model. This process is based on the definition of step S1 to achieve risk measurement, analysis and processing of Internet-related activity data.
[0068] The following will take online speech data as an example to illustrate the processing steps, and step S2 includes the following specific steps:
[0069] S21. Internet data access: First, through regular data synchronization processing, Internet data / Internet activity data such as online speech, communication information, transaction shopping, logistics and delivery are acquired and aggregated.
[0070] S22, pre-processing the collected data. Step S22 includes the following steps:
[0071] S221. Internet-related behavior matching processing, i.e. identifying the type of Internet-related data, and selecting corresponding processing rules for the accessed Internet-related activity data according to its data type. The processing rules include scoring method and weighted processing method.
[0072] S222. After determining that the data type is the type of online speech, continue with standardization processing, that is, identify the data columns such as "speech account, speech time, speech content" in this type of data and store them in the database according to the specifications, then select the scoring method (here it is selected to score by the number of hits), and select the weighted processing method (detailed description in the following steps).
[0073] S23, data content extraction: batch read speech content, speech account and other data columns. Extract the standardized data to obtain the text data to be analyzed.
[0074] S24, loading basic library: loading the basic library information defined in step S21, including: vocabulary library, formula library, and rule library.
[0075] S25, score superposition calculation and weighted processing. According to the matched rules in step S22 and the basic library loaded in step S24, the data is analyzed and scored, and additional points and weighted processing are performed, and the scores are accumulated to the relevant network accounts.
[0076] Score superposition operation: compare and match the speech content extracted from the data content of step S23, all sensitive words and civilian substitute words according to the categories to which the sensitive words belong, and score the hit sensitive words and corresponding civilian substitute words according to the sensitive word categories to which they belong. Because the scores of sensitive words in different categories are different, they need to be scored separately.
[0077] In step S25, the score superposition operation includes the following steps:
[0078] S251. Online speech data must be processed according to the “number of hits”. Therefore, when a sensitive word or alternative word is hit for the first time, the corresponding score is added according to the category to which these words belong.
[0079] S252. When the same sensitive word is hit again, points will continue to be added.
[0080] S253. When other sensitive words in the same category are hit, points are added continuously in the same manner as steps S251 and S252.
[0081] S254: If multiple hit sensitive words in the same category match the formula library, additional scoring is performed according to the corresponding formula rules.
[0082] S255. When a sensitive word in a different category is hit, in addition to scoring the sensitive word, an additional scoring process is performed.
[0083] S256. When the same data also hits the exclusion word, no points are counted or points are reduced according to the type of the exclusion word hit.
[0084] S257. Finally, the total score of the data is obtained, and the score is added to the relevant speaking account.
[0085] S258, repeating steps S251 to S257 until all data are traversed. The data includes online comments, communication information, transaction shopping, and logistics delivery.
[0086] In step S25, weighted processing includes: for the network behavior involving a certain account, if it involves multiple data types (such as online speech, transaction shopping) at the same time and hits sensitive words or civilian substitute words under the data type, it is necessary to perform weighted average calculation on the total score of the account (set the proportion of each data type in the total score), and then perform additional score processing on the total score.
[0087] S26, obtain and output the risk index of the account involved in the online activities, and output the online activity risk index of the account (including the risk level and the total score). Finally, the total score of the network accounts related to these network activities is obtained, and converted into a risk level (high, medium, low), and the online crime risk index of the network account is output.
[0088] The present invention is a text content detection method for network activity data, which combines related services, text analysis technology and custom rule engine technology, and integrates the concept of points to calculate the network-related crime risks of related behaviors. The present invention can support the risk detection business of crimes involving explosions and gun trafficking.
[0089] The present invention can detect text data of online activities such as online speech, communication information, transaction shopping, logistics delivery, etc., obtain the risk index of related accounts through comprehensive calculation, and can quickly and objectively analyze the target account, thereby alleviating the workload of staff who need to manually check network accounts and their activity behaviors one by one, reducing the difficulty of manual analysis and judgment, and allowing staff to quickly predict the account risks involved in the behavior.
[0090] The present invention is based on a customized and configurable rule engine, which fully considers relevant business scenarios, designs integral calculation rules for different scenarios, and more deeply integrates into the explosion-related and gun trafficking businesses, and designs different explosive formula rules, material and accessory combination rules. The rule engine can be adjusted at any time according to changes in future staff needs. It can not only quickly meet new needs, but also reduce the development and maintenance costs of related systems.
[0091] The present invention adopts the concept of account risk management and control, and adopts a point calculation mechanism for the detection of Internet-related activities. According to various rules set by the rule engine, combined with weighted processing, points are calculated centered on relevant network accounts. The result of the calculation is a comprehensive score, which is converted into a corresponding risk level. Compared with the past, when staff could only receive scattered clues of suspicious Internet-related criminal activities one by one, the present invention allows staff to identify high-risk and high-score accounts at a glance, and can check related activity data, eliminating the work of manual verification and manual analysis by staff.
[0092] The present invention combines the business of related departments, establishes a related vocabulary, sorts out exclusion words according to multiple dimensions such as different occupations, units, and places, establishes an exclusion mechanism, reduces noise, and reduces the push of useless information to staff.
[0093] The entire analysis method of the present invention adopts an automated processing mechanism. As long as there is continuous access to network activity data, various types of network-related activity data can be automatically analyzed, and high-risk network-related criminal activities and related network accounts can be discovered.
[0094] The present invention is a method for calculating, detecting and analyzing the risks of Internet-related crimes based on Internet activity data. It is a new method based on the integration of multiple technologies such as big data computing, self-developed rule engines and text analysis. It can not only quickly respond to changes in business needs, but also make the prompts of Internet-related crime risks more intuitive. It can also set relevant scoring and exclusion methods through a flexible rule engine to make detection operations more automatic and accurate.
Claims
1. A method for intelligently revealing the risks of Internet-related crimes, characterized in that: By monitoring and classifying the text content of relevant network activity data, combined with the defined basic library, using rule engines, big data comparison, and text analysis technology, relevant network activity data is extracted, analyzed, and integrated, and the relevant scores of the series of network activities are output. The points are accumulated to the relevant network accounts, and the final risk level of the account is obtained after risk level conversion; The specific steps include: S1. Conduct basic construction, including building a vocabulary, building an explosive formula library, and formulating detection rules; In step S1, constructing the vocabulary includes the following steps: S101. Integrate the relevant investigation experience of the public security department and combine it with the actual business of explosive-related criminal activities to sort out and establish a sensitive word database for online activities related to explosives, and classify sensitive words; S102. Sort out and collect civilian alternative explosive materials, establish a civilian alternative word library, and set a one-to-one correspondence with the sensitive word library of online activities related to explosives; S103, sort out the activities related to explosive making but not suspected of crime, establish a corresponding exclusion word library, and set the exclusion method of each exclusion word; S2. Define the "integrated Internet-related crime risk disclosure process" model and start the "integrated Internet-related crime risk disclosure process" model to perform risk measurement and processing on Internet-related activity data; The step S2 comprises the following specific steps: S21. Internet data access: obtaining and aggregating Internet data; S22, data preprocessing; S23, data content extraction; S24, loading basic library: loading the basic library information defined in step S21; S25, score superposition calculation and weighted processing; Score superposition operation: compare and match the content extracted from the data content in step S23 according to the category to which the sensitive words belong, and score the hit content according to the category to which it belongs; Weighted processing: For online behaviors involving a certain account, if multiple data types are involved and sensitive words or civilian substitute words under the data type are hit, the total score of the account needs to be weighted averaged and then additional points are added to the total score; In step S25, the score superposition operation includes the following steps: S251. Online speech data must be processed according to "hit count"; S252: If the same sensitive word is hit again, the score will continue to increase; S253. If other sensitive words in the same category are hit, points will continue to be added; S254. If multiple hit sensitive words in the same category match the formula library, additional scoring is performed according to the corresponding formula rules; S255. When a sensitive word in a different category is hit, in addition to scoring the sensitive word, an additional scoring process is performed; S256, when the same data also hits the exclusion word, no score or score reduction is performed according to the type of the exclusion word hit; S257, finally obtaining the total score of the data, and adding the score to the relevant speech account; S258, repeat step S251 to step S257 until all data are traversed; S26. Obtain and output the risk index of the account involved in Internet activities.
2. According to claim 1, a method for intelligently revealing the risks of Internet-related crimes is characterized in that: In step S1, constructing an explosive formula library includes the following steps: sorting out the raw materials required for the explosive formula according to the preparation requirements of each type of explosive, establishing a relevant formula table for each type of explosive, and establishing an explosive formula library.
3. According to claim 1, the method for intelligently revealing the risks of Internet-related crimes is characterized in that: In step S1, formulating detection rules includes the following steps: S111. Formulate scoring rules. According to different types of characteristics, the scoring rules are divided into two categories: "scoring by number of hits" and "scoring by number / weight of hits"; S112. Formulate formula rules. According to the raw material requirements for explosives, after sorting out and establishing the relevant formula library, formulate relevant weighted operation rules for each formula.
4. According to claim 1, a method for intelligently revealing the risks of Internet-related crimes is characterized in that: The step S22 comprises the following steps: Identify the types of Internet-related data, perform standardized data processing and storage on this type of data, and match the data with corresponding rules.
Citation Information
Patent Citations
Calculation method of crime degree of speech data
CN105138570A
A data-driven network sensitive information mining and early warning platform
CN109739849A
Sensitive account detection method, storage medium and computer equipment
CN111079029A
Network crime platform identification method, system and device and computer storage medium
CN113407886A
Server and method for managing crime using big data
KR1020160001994A
