Natural person identity identity discrimination method

Abstract

The invention discloses a natural person identity identity discrimination method, which comprises two dimensions: dimension 1: a seed terminal identifier has direct identity information association in data, or the seed terminal identifier is directly associated with identity document information on a data level; dimensionality 2: the seed terminal identifier cannot be associated with the identity information on the data level; during calculation, the confidence coefficient of a new user is improved in the aspect of time dimension, meanwhile, according to a same-name triangle verification algorithm, a name which can be extracted by a telecommunication terminal code address in real time and a name corresponding to an identity code address are verified, if triangle verification is passed, the confidence coefficient of the identity code address is improved again, and through various model algorithms, the identity code address can be identified. The real identity of the target person is excavated layer by layer.

Description

Technical Field

[0001] The present invention belongs to the field of communication technology, and in particular relates to a method for identifying the identity of a natural person. Background Art

[0002] With the advent of the information age, all aspects of people's lives have begun to connect with the Internet, become digital and virtual. In the era of big data, mining based on Internet data can often bring more added value. In the field of national security, due to the virtuality, transmission, freedom and openness of the Internet, the maintenance of the Internet environment has become complicated. Internet fraud, foreign espionage and other incidents are common. The relationship between a person's multiple identities on the Internet and the real identity, the relationship between virtual accounts and telecommunications terminal identifiers, the complex relationship between historical users and current users, and the current huge amount of data and the chaos of the Internet all pose more arduous challenges to identifying a person's true identity in real time. At present, the industry mainly relies on the big data platform to combine various types of data crawled from the Internet for research and judgment. The main technical feature is to label the data, assign different confidence levels to different data sources based on historical experience, and finally comprehensively calculate the identity information and feature information of the actual user of the seed code address.

[0003] Although previous technologies can infer the real identity information of the actual user based on the combined effect of multiple data sources, they are quite limited. In the case of few data sources, the error is large. At the same time, errors may occur in certain specific situations. For example, if the real identity information of the current actual user is inferred from a seed code address, and the original user of this seed code address has been cancelled, and the new user has just registered the seed code address, this scenario will cause the inferred real identity information of the actual user to be the previous user. The reason for this problem is that the original user generates more data than the new user, and the confidence of the historical user calculated by the confidence hybrid algorithm based on multiple data sources will be higher. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the present invention provides a method for identifying the identity of a natural person, including two dimensions. Dimension one: the seed terminal identifier has a direct identity information association in the data, or the seed terminal identifier is directly associated with the identity document information at the data level; Dimension two: the seed terminal identifier cannot be associated with the identity information at the data level; when calculating, the present invention improves the confidence of the new user in the time dimension, and at the same time, based on the same-name triangular verification algorithm, verifies the name that can be extracted in real time by the telecommunications terminal code address and the name corresponding to the identity code address. If the triangular verification passes, the confidence of the identity code address will be further improved, and the true identity of the target person will be dug out layer by layer through various model algorithms.

[0005] The technical solution adopted by the present invention to solve the technical problem is as follows:

[0006] Step 1: The identification of the identity of the natural person is divided into two dimensions: Dimension 1: the seed terminal identifier is directly associated with the identity information in the data, or the seed terminal identifier is directly associated with the identity document information at the data level; Dimension 2: the seed terminal identifier cannot be associated with the identity information at the data level;

[0007] Step 2: The dimension 1 includes two models: a multi-data source confidence hybrid algorithm and a same-group confidence hybrid algorithm;

[0008] Step 2-1: Multiple data source confidence hybrid algorithm;

[0009] 1) Data source dimension;

[0010] The confidence level of the expansion mark is related to the data source to which it belongs. The data source is marked and scored based on past experience, and different data sources contribute different scores.

[0011] The more associated data sources there are, the higher the score will be;

[0012] 2) Time dimension;

[0013] Differentiate between dynamic data and static data. Dynamic data is sorted in reverse order based on update time. The later the update time, the higher the score.

[0014] 3) Number of data sources;

[0015] For flow behavior data sources, the more associations there are, the higher the score will be;

[0016] There are two dimensions: time dimension and data source. The time dimension is higher than the number of data sources.

[0017] The number of non-communication dynamic associations is the same at the same time, but the number of times is different;

[0018] 4) Prior knowledge;

[0019] ① The same type of logos in the expansion result exceed maxFilterSize, and it is determined to be a counterfeit mobile phone;

[0020] ②Names and nicknames are not allowed to be linked;

[0021] ③The copycat IMEI does not expand its line;

[0022] ④The default level for line expansion is 3 layers;

[0023] ⑤The default time for line expansion is nearly one year;

[0024] ⑥The maximum number of expansion line logos is 100;

[0025] ⑦ The length of the expansion line mark is less than 50;

[0026] 5) Algorithm implementation;

[0027] Through the hybrid calculation model of multi-data source confidence, the confidence of the data source and the clustering and scoring algorithm of each identity code address information, the scores of each identity code address are finally converted into confidence for sorting;

[0028] c=p1*m1+m1*α*(d1+d2+…+dn)+β

[0029] Among them, c is the confidence, p1 is the parent identifier confidence, m1 is the level confidence, d1, d2, ... are the basic confidences of each data source;

[0030] First layer: 0.8, second layer: 0.7, third layer: 0.6, fourth layer: 0.5; the initial input mark is 0 layer, the mark confidence is 1, and it is not included in the above formula calculation;

[0031] α: confidence coefficient of data source dimension;

[0032] γ: confidence coefficient of time dimension;

[0033] β: confidence parameters of time and data source frequency dimensions. For different identifiers of the same type, confidence parameters of time dimension and data source frequency are added.

[0034] β=γ*m1*(size-n)

[0035] size: the number of different identifiers of the same type and at the same layer;

[0036] n: sorting sequence number;

[0037] Step 2-2: Hybrid algorithm for identity code and address confidence of the same group;

[0038] As the telecommunication terminal code values ​​used by members of the same group are all opened by one of the members, a hybrid algorithm for the confidence of the identity code address of the same group is designed, as follows:

[0039] 1) Input the seed telecommunication terminal code value;

[0040] 2) Query the seed telecommunications terminal code value to obtain the identity code address information actually associated with the data;

[0041] 3) Obtain the identity code information of other members in the same group based on the identity code information;

[0042] 4) Obtain the corresponding name based on the identity code address information of other members;

[0043] 5) Triangle verification with the same name;

[0044] Step 3: The second dimension includes two models: a feature path-relationship collision algorithm and a feature path-behavior information collision algorithm model;

[0045] Step 3-1: Feature path-relationship collision algorithm;

[0046] When the seed code address cannot be directly linked to the identity code address information at the data level, the identity information of the actual user can be determined through the first-level contact; process

[0047] 1) Input the seed telecommunication terminal code value;

[0048] 2) First-level contact person who queries the seed telecommunications terminal code value;

[0049] 3) Query the identity code address associated with the telecommunication terminal identification of the first-level contact;

[0050] 4) Query the information of other members in the same group based on the identity code address set, and obtain the identity code addresses of other members in the same group;

[0051] 5) Query the identity code address set of other members of the same group to obtain the name corresponding to the identity code address;

[0052] 6) Triangle verification with the same name;

[0053] Step 3-2: Feature pathway-behavior information collision algorithm;

[0054] The algorithm relies on the data source of travel information records and the dynamic text information received by the telecommunication terminal code value to extract content;

[0055] 1) Enter the telecommunication terminal code value;

[0056] 2) Extracting travel information and name from recent text content information based on the seed telecommunication terminal code value;

[0057] 3) Based on the extracted travel information, search the travel records of the entire network, find the records that match the text content information, and extract the corresponding identity code address information and name;

[0058] 4) Triangle verification with the same name.

[0059] Preferably, the flow behavior data source includes the user's communication behavior and web browsing behavior.

[0060] Preferably, the non-communication dynamic association times include multiple identity code address information corresponding to the telecommunication terminal identifier in the travel behavior data source.

[0061] Preferably, the default value of α is 0.01.

[0062] Preferably, the default value of γ is 0.01.

[0063] Preferably, the confidence level of the characteristic path-relationship collision algorithm is: c=0.83.

[0064] Preferably, the collision confidence of the characteristic path-travel information collision algorithm is: c=0.83.

[0065] The beneficial effects of the present invention are as follows:

[0066] Compared with the previous hybrid confidence algorithm, the technology of the present invention adds more business model algorithms, such as same-name triangle verification, time dimension algorithm, relationship collision model, feature path model, etc., and these models are used to verify each other with the traditional multi-data source hybrid confidence algorithm, so that the identity identification of natural persons is more accurate. For example, the actual identity information of the current actual user is calculated from a certain seed code address, and it happens that the original user of this seed code address has been cancelled, and the new user has just registered the seed code address. This scenario will cause the actual identity information of the actual user calculated to be the previous user. The reason for this problem is that the data generated by the original user is more than the data of the new user, and the confidence of the historical user calculated by the confidence hybrid algorithm based on multiple data sources will be higher. When calculating, the algorithm of the present invention will increase the confidence of the new user in the time dimension, and at the same time, according to the same-name triangle verification algorithm, the name that can be extracted in real time by the telecommunications terminal code address is verified with the name corresponding to the identity code address. If the triangle verification passes, the confidence of the identity code address will be increased again, and the real identity of the target person will be excavated layer by layer through various model algorithms. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 is a schematic diagram of the microfluidic device of the present invention; Figure 2 It is a flow chart of the hybrid algorithm of the group identity code address confidence of the present invention; Figure 3 It is the relationship collision flow chart of the present invention; Figure 4 It is a travel collision flow chart of the present invention. DETAILED DESCRIPTION

[0068] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0069] The ultimate goal of the present invention is to start from the original seed code address, expand the associated code addresses one level at a time, and based on a specific model algorithm, dig out the actual identity information of the actual user of the seed code address layer by layer. However, the existing implementation schemes have more or less deficiencies, which leads to deviations in the natural identity identification of the actual user. In response to the above deficiencies, this technical solution, based on long-term professional business experience, extracts more algorithm models to support each other, combines practice with theory, and makes natural person identity identification more objective and accurate.

[0070] The present invention uses known seed terminal identifiers and more business models to verify each other when inferring the target identity, enriching the natural person identity recognition algorithm and making the natural person identity recognition more objective and accurate. Figure 1 Seed code address mining target identity flow chart, the specific implementation process is as follows:

[0071] Identity identification can be generally divided into two dimensions. First, the seed terminal identifier has a direct identity information association in the data, or the seed terminal identifier can be directly associated with the identity document information at the data level; second, the seed terminal identifier cannot be associated with the identity information at the data level. The following is a detailed introduction to the algorithm implementation process of the two-dimensional identity identification model.

[0072] Dimension one includes two models: a multi-data source confidence hybrid algorithm and a same-group confidence hybrid algorithm.

[0073] The multi-data source confidence hybrid algorithm is designed with the following dimensions:

[0074] 1) Data source dimension;

[0075] The confidence level of the expansion line identification is closely related to the data source to which it belongs. The data source needs to be labeled and scored based on past experience, and different data sources contribute different scores.

[0076] The more associated data sources there are, the higher the score.

[0077] 2) Time dimension;

[0078] Differentiate between dynamic data and static data, and sort dynamic data in reverse order based on update time. The later the update time, the higher the score.

[0079] 3) Number of data sources;

[0080] Flow behavior data sources, such as user communication behavior, web browsing behavior and other data, the more times they are associated, the higher the points.

[0081] There are two dimensions: time dimension and data source. The time dimension is higher than the data source number. Reason: The original address user affects the inference of the current user.

[0082] The number of non-communication dynamic associations, for example, when the telecommunications terminal identifier in the travel behavior data source corresponds to multiple identity code address information, the time is the same but the number of times is different.

[0083] 4) Prior knowledge;

[0084] ① As a result of line expansion, the number of similar identifiers exceeds maxFilterSize (adjustable). If the maxFilterSize value is 10, identifiers with more than 10 similar identifiers will be discarded. If the seed terminal code address is associated with more than 10 IMEIs (International Mobile Equipment Identity), it can be considered a counterfeit mobile phone.

[0085] ② Names and nicknames are not linked.

[0086] ③The copycat IMEI does not expand its line.

[0087] ④ The default level of line expansion is 3 (three levels can associate all different identification types of the target person).

[0088] ⑤The default time for line expansion is nearly one year.

[0089] ⑥ The maximum number of expansion line logos is 100 (experience shows that the number of human logos generally does not exceed 100).

[0090] ⑦ The length of the expansion line mark is less than 50 (to prevent illegal data).

[0091] 5) Algorithm implementation:

[0092] Through the hybrid calculation model of multi-data source confidence, data source confidence and clustering scoring algorithm for each identity code address information, finally, the scores of each identity code address are converted into confidence for sorting.

[0093] c=p1*m1+m1*α*(d1+d2+…+dn)+β

[0094] First layer: 0.8, second layer: 0.7, third layer: 0.6, fourth layer: 0.5; the initial input identification is 0 layer, the identification confidence is 1, and it is not involved in the calculation of the above formula.

[0095] For example, the telecommunications terminal identifier msisdn-138 is expanded to obtain two network virtual identifiers qq, qq1-4617, qq2-4618, qq1 updated on 20190401, data source times 200, qq2 updated on 20190302, data source times 500, update time as the first sorting reference, data source times as the second sorting reference. Then qq1's β = α*(size-n) = 0.01*0.8*(2-1) = 0.008, qq2's β = α*(size-n) = 0.01*0.8*(2-2) = 0.

[0096] A hybrid algorithm for the confidence of the identity code address of the same group;

[0097] There is a very common situation at present that the telecom terminal code values ​​used by members of the same group are all opened by one of the members. This situation leads to the fact that the identity information of the real user can never be obtained based on the data fact level. An algorithm is designed for this situation. Reference Figure 2 A hybrid algorithm for the confidence of the identity code address of the same group;

[0098] 1) Input the seed telecommunication terminal code value;

[0099] 2) Query the seed telecommunications terminal code value to obtain the identity code address information actually associated with the data;

[0100] 3) Based on the identity code address set obtained in the second step, query and obtain the identity code address information of other members in the same group;

[0101] 4) Obtain the corresponding name based on the identity code address obtained in the third step;

[0102] 5) Triangle verification with the same name;

[0103] The confidence algorithm of the relationship collision package algorithm is:

[0104] c=0.83

[0105] 0.83 The reasons are as follows:

[0106] When the input seed telecommunication terminal code value is expanded to the identity code address, this identity code address is verified by 4 data sources (and the confidence of the 4 data sources is 0.9), then the confidence is about 0.8288. When the data sources are greater than 4, the confidence of this mixed confidence algorithm is higher than the confidence of the relational collision group card algorithm. The same applies to 0.83.

[0107] Dimension 2 includes two models, the feature path-relationship collision algorithm and the feature path-behavior information collision algorithm model. The seed code address cannot be directly associated with the identity code address at the data fact level. We call the algorithm that opens the information path from the seed code address to the identity code address "feature path" algorithm. This dimension has the following two solutions.

[0108] Feature Path-Relation Collision Algorithm;

[0109] When the seed code address cannot be directly linked to the identity code address information at the data level, the identity information of the actual user can be determined through the first-level contact. Figure 3 Relationship collision flow chart;

[0110] 1) Enter the seed telecommunications terminal code value.

[0111] 2) First-level contact to query the seed telecommunications terminal code value.

[0112] 3) Query the identity code address associated with the telecommunications terminal identification of the first-level contact.

[0113] 4) Query the information of other members in the same group based on the identity code address set, and obtain the identity code addresses of other members in the same group.

[0114] 5) Query the identity code set of other members of the same group to obtain the name corresponding to the identity code.

[0115] 6) Triangle verification with the same name.

[0116] The confidence level of the “feature path”-relationship collision is:

[0117] c=0.83

[0118] Feature Path-Behavior Information Collision Algorithm:

[0119] The algorithm relies on the data source of travel information records and the dynamic text information extracted from the telecommunications terminal code value recently received. According to the rules, it is very unlikely that the same name will travel on the same day and take the same bus. Figure 4 Behavior Information Collision Flowchart

[0120] 1) Enter the telecommunications terminal code value.

[0121] 2) Extract travel information and name from recent text content information based on the seed telecommunication terminal code value.

[0122] 3) Based on the travel information records extracted in the second step, search the travel records of the entire network, find the records that match the text content information, and extract the corresponding identity code address information and name.

[0123] 4) Triangle verification with the same name.

[0124] The confidence level of the “feature path”-behavior information collision is:

[0125] c = 0.83.

Claims

1. A method for identifying the identity of a natural person, characterized in that: The steps include: Step 1: The identification of the identity of the natural person is divided into two dimensions: Dimension 1: the seed terminal identifier is directly associated with the identity information in the data, or the seed terminal identifier is directly associated with the identity document information at the data level; Dimension 2: the seed terminal identifier cannot be associated with the identity information at the data level; Step 2: The dimension 1 includes two models: a multi-data source confidence hybrid algorithm and a same-group confidence hybrid algorithm; Step 2-1: Multi-data source confidence hybrid algorithm; 1) Data source dimension; The confidence level of the expansion mark is related to the data source to which it belongs. The data source is marked and scored based on past experience, and different data sources contribute different scores. The more associated data sources there are, the higher the score will be; 2) Time dimension; Differentiate between dynamic data and static data, and sort dynamic data in reverse order based on update time; The later the update time, the higher the points; 3) Number of data sources; For flow behavior data sources, the more associations there are, the higher the score will be; There are two dimensions: time dimension and data source. The time dimension is higher than the number of data sources. The number of non-communication dynamic associations is the same at the same time, but the number of times is different; 4) Prior knowledge; ① The same type of logos in the expansion result exceed maxFilterSize, and it is determined to be a counterfeit mobile phone; ②Names and nicknames are not allowed to be linked; ③The copycat IMEI does not expand its line; ④The default level for line expansion is 3; ⑤The default time for line expansion is nearly one year; ⑥The maximum number of extension line marks is 100; ⑦ The length of the expansion line mark is less than 50; 5) Algorithm implementation; Through the hybrid calculation model of multi-data source confidence, the confidence of the data source and the clustering and scoring algorithm of each identity code address information, the scores of each identity code address are finally converted into confidence for sorting; c=p1*m1+m1*α*(d1+d2+…+dn)+β Among them, c is the confidence, p1 is the parent identifier confidence, m1 is the level confidence, d1, d2, ... are the basic confidences of each data source; First layer: 0.8, second layer: 0.7, third layer: 0.6, fourth layer: 0.5; the initial input mark is 0 layer, the mark confidence is 1, and it is not included in the above formula calculation; α: confidence coefficient of data source dimension; γ: confidence coefficient of time dimension; β: confidence parameters of time and data source frequency dimensions. For different identifiers of the same type, confidence parameters of time dimension and data source frequency are added. β=γ*m1*(size-n) size: the number of different identifiers of the same type and at the same layer; n: sorting sequence number; Step 2-2: Hybrid algorithm for identity code and address confidence of the same group; As the telecommunication terminal code values ​​used by members of the same group are all opened by one of the members, a hybrid algorithm for the confidence of the identity code address of the same group is designed, as follows: 1) Input the seed telecommunication terminal code value; 2) Query the seed telecommunications terminal code value to obtain the identity code address information actually associated with the data; 3) Obtain the identity code information of other members in the same group based on the identity code information; 4) Obtain the corresponding name based on the identity code address information of other members; 5) Triangle verification with the same name; Step 3: The second dimension includes two models: a feature path-relationship collision algorithm and a feature path-behavior information collision algorithm model; Step 3-1: Feature path-relationship collision algorithm; When the seed code address cannot be directly linked to the identity code address information at the data level, the identity information of the actual user can be determined through the first-level contact; process 1) Input the seed telecommunication terminal code value; 2) First-level contact person who queries the seed telecommunications terminal code value; 3) Query the identity code address associated with the telecommunication terminal identification of the first-level contact; 4) Query the information of other members in the same group based on the identity code address set, and obtain the identity code addresses of other members in the same group; 5) Query the identity code address set of other members of the same group to obtain the name corresponding to the identity code address; 6) Triangle verification with the same name; Step 3-2: Feature pathway-behavior information collision algorithm; The algorithm relies on the data source of travel information records and the dynamic text information received by the telecommunication terminal code value to extract content; 1) Enter the telecommunication terminal code value; 2) Extracting travel information and name from recent text content information based on the seed telecommunication terminal code value; 3) Based on the extracted travel information, search the travel records of the entire network, find the records that match the text content information, and extract the corresponding identity code address information and name; 4) Triangle verification with the same name.

2. A method for identifying the identity of a natural person according to claim 1, characterized in that: The flow behavior data source includes the user's communication behavior and web browsing behavior.

3. A method for identifying the identity of a natural person according to claim 1, characterized in that: The non-communication dynamic association times include multiple identity code address information corresponding to the telecommunication terminal identifier in the travel behavior data source.

4. A method for identifying the identity of a natural person according to claim 1, characterized in that: The α defaults to 0.

01.

5. A method for identifying the identity of a natural person according to claim 1, characterized in that: The default value of γ is 0.

01.

6. A method for identifying the identity of a natural person according to claim 1, characterized in that: The confidence level of the characteristic path-relationship collision algorithm is: c=0.

83.

7. A method for identifying the identity of a natural person according to claim 1, characterized in that: The collision confidence of the characteristic path-travel information collision algorithm is: c=0.83.

Images