CN108268762B - Mobile social network user identity identification method based on behavior modeling - Google Patents
Mobile social network user identity identification method based on behavior modeling Download PDFInfo
- Publication number
- CN108268762B CN108268762B CN201810043919.7A CN201810043919A CN108268762B CN 108268762 B CN108268762 B CN 108268762B CN 201810043919 A CN201810043919 A CN 201810043919A CN 108268762 B CN108268762 B CN 108268762B
- Authority
- CN
- China
- Prior art keywords
- user
- text
- data
- information
- behavior
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
- G06F21/31—User authentication
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Abstract
The invention researches a mobile social network user identity identification method based on behavior modeling. Therefore, the consistency of the user behaviors is judged, and the legality of the user is judged according to the consistency. The mobile social network user identity identification method based on behavior modeling benefits from rich information data provided by the current mobile social network, and obtains a large amount of user social data including positions, mobile modes, social relations, user generated contents, shopping records and the like. And taking the multivariate data as a research object, and extracting the characteristics of the information. The modeling of the user behavior in the following three dimensions is realized by adopting methods such as mixed kernel density estimation, LDA text modeling and the like: (1) the user at the geographic position (2) generates a text (3) and the social relationship is calculated by simulating a user identity embezzlement experiment, so that the performance of the system is evaluated, the defects of the traditional identity recognition system are solved, and a new thought and an analysis method are provided for solving the safety problem in the information era.
Description
Technical Field
The invention relates to a method for identifying the identity of a mobile social network user.
Background
While the internet application is popularized, the problem of information security is more and more prominent, and an attacker can try to steal a user account, steal personal privacy information and even maliciously attack a server. In order to ensure the safety of users, the traditional identity identification method is very easy to be stolen under the increasingly intensified network attacks, so that the actual requirements cannot be met only by a simple system of user names and passwords.
Today, internet intellectualization is a development trend, identity sensitive information of a user can be widely collected and used, and how to ensure information security of the user becomes a difficult problem to be solved urgently. Compared with the prior art, the method has the advantages of uniqueness, irreproducibility, non-repudiation and the like by taking the user inertial behavior characteristics as the basis of identity identification, has very high accuracy as long as the method is properly designed, does not depend on other auxiliary equipment, does not interfere with the normal application of the user, can realize the authentication of the user identity only in a plug-in mode, and can accurately monitor the stealing of the user identity in real time. The method comprises the steps that a new visual angle is provided for mobile internet information safety problems based on user identity identification problems discovered by abnormal behaviors, and a user daily geographic position distribution model can be obtained by utilizing a check-in place of a user according to a user check-in record and by utilizing kernel density estimation modeling by utilizing mass information generated in a mobile social network and considering the characteristics of user behaviors in a physical-network-social space; extracting user interest characteristics by utilizing a text modeling mode widely applied in data mining, wherein the user interest characteristics can be used as behavior data for identifying the identity of a user; and establishing a user social preference model by utilizing the user social relationship, and the like.
An important challenge often faced in modeling mobile social network user behavior is the data sparsity problem. Due to various condition limitations, collected records of specific behaviors of the user are often sparse, and the collected records are important factors influencing the modeling accuracy of the user behaviors. We propose to complete the modeling of user behavior by filling the spatiotemporal holes of behavior data based on complementary effects of user behavior projections (user behaviors in different behavior subspaces).
Therefore, the invention researches a mobile social network user identity identification method based on behavior modeling. Therefore, the consistency of the user behaviors is judged, and the legality of the user is judged according to the consistency.
Disclosure of Invention
Due to the rich information data provided by the current mobile social network, a large amount of user social data including positions, mobile modes, social relations, user generated contents, shopping records and the like can be obtained. And taking the multivariate data as a research object, and extracting the characteristics of the information. The modeling of the user behavior in the following three dimensions is realized by adopting methods such as mixed kernel density estimation, LDA text modeling and the like: (1) the method comprises the steps that a user at a geographic position (2) generates a text (3) social relation, the detected interception rate, disturbance rate and precision rate are calculated through a simulated user identity stealing experiment, the system performance is evaluated, a mobile social network user behavior identity identification method based on abnormal discovery is designed, the defects of a traditional identity identification system are overcome, and a new thought and analysis method is provided for solving the information age safety problem.
The method and the device overcome the defects of the prior art, are used for analyzing the consistency of the user behavior and the threshold value in the expected model, and research whether the user account is abnormal in the mobile social network.
Therefore, the technical scheme is as follows:
a user identity false-identification algorithm based on behavior modeling is characterized by comprising the following processes:
inputting: user-generated check-in information, text information, and social relationships
And (3) outputting: truth value of user identity
(1) Calculating the characteristic distribution rule of the user historical data in three dimensions according to the user check-in geographical position information, the text content sent by the user online behavior space and the social relationship of the user in the whole social network
And
executing the step (2);
(2) for a newly generated behavior record of a user, it is evaluated in three dimensions: calculating a log mean of probability density values of newly signed data
Calculating a topic probability distribution for newly generated text
Calculating interest distribution for newly established social relationships
The method specifically comprises the following steps:
obtaining probability density function f (x) of sign-in position distribution by adopting a check density estimation method according to historical sign-in data of users and friends of the users, substituting the data of the new sign-in of the users into the f (x) to obtain probability density, and calculating the mean logarithmic value of the probability density function
The above-mentioned
Obtaining topic probability distribution of new text
Where n (k) is the number of words in the text belonging to topic k, αkModel parameters obtained in the LDA training process; the interest distribution of the new friend is obtained by extracting the text data generated by the new friend in the same way
Executing the step (3);
(3) in the check-in information dimension, to be calculated
The value of (A) and a probability density logarithmic mean threshold value S obtained by training and capable of distinguishing authenticity of user identity through geographic positioncComparing, if greater than threshold, returning UCOtherwise, return UCExecuting the step (4) as false;
(4) subject probability distribution for user historical data in the text information dimension
And subject probability distribution of user newly generated text
Jensen-Shannon divergence calculating two probability distributions
The value of the JS divergence is compared with a threshold value D of JS divergence between two probability distributions obtained through trainingTComparing, if less than threshold, returning UTOtherwise, return UTExecuting the step (5) as false;
(5) in societyAnd (4) cross relationship dimension, extracting text information capable of expressing interest in friend information according to social relationship of the user, and calculating the probability distribution of the topics of the friend historical data of the user in the same way
And subject probability distribution of user newly generated text
And JS divergence
Its value is compared with a threshold value DFComparing, if less than threshold, returning UTOtherwise, return UTExecuting step (6) as false;
(6) the judgment U for the user identity truth value returned according to the steps (3), (4) and (5)C，UT，UFThe judgment of the authenticity of the user identity in three dimensions is combined to obtain the final judgment value U of the user identityI＝g(UC，UT，UF)。
Firstly, based on the social network user behavior rules of the position, a user normal behavior model is established in multiple dimensions by utilizing an effective data set, feature selection is carried out, an abnormal behavior threshold value is determined, and the accuracy and the effectiveness of the abnormal behavior threshold value are verified through experiments. And then, based on the previous research on user behavior feature extraction and normal behavior modeling, analyzing the features of the user in a network space by combining offline behavior, online behavior and social behavior, comprehensively considering the relevance of a user behavior synthesis space, and designing a user identity identification method and system realized by utilizing the complementary effect of multiple dimensions.
Experiments prove that the method is superior to the prior research in both accuracy and calculation time.
Drawings
FIG. 1 is a system structure diagram of a mobile internet user identification method based on behavior modeling
FIG. 2 is a flow chart of the algorithm of the present invention
Detailed Description
(case)
A system structure diagram of a mobile internet user identity authentication method based on behavior modeling is shown in fig. 1. The whole scheme is divided into three stages:
the method comprises the steps that in the first stage, a user offline model is established for historical data, and is responsible for collecting and processing the historical data of a user to generate a distribution rule of user behavior characteristics of corresponding dimensions; (as is conventional in the art)
The second stage is a stage of collecting behavior records on line to generate current characteristics, and is responsible for collecting and processing current data of a user, and calculating to obtain a distribution rule of behavior characteristics of corresponding dimensions; (as is conventional in the art)
The third stage is a user identity identification stage which is responsible for identifying the false according to the data characteristics delivered in the first stage and the second stageMultiple purpose Dimension fusionA determination of the identity of the user is given.
The first stage comprises the following concrete implementation steps:
step 1-1, preprocessing user data and screening effective users of each dimension. The conditions valid for the three dimensions are: the user history check-in record needs to have more than 5 check-in records; the text content sent by the user in the online behavior space must be sufficient enough, and the theme probability distribution of the user historical data can be obtained through training, namely the user who removes the stop word and ensures more than 200 effective words is ensured; the number of friends of the user in the whole network is more than ten, and based on the characteristics, the text content of the friends of the user is recorded, so that the interest distribution of a friend circle in the user history data is obtained.
Step 1-2, according to historical check-in data of the user and friends of the user, a check-in density estimation method (which is the prior art in the field) is adopted to obtain a probability density function f (x) of check-in position distribution.
And 1-3, accumulating the historical text data of each user as a document, wherein the text data of all the users form a large-scale corpus. Establishing a theme model for the historical text data of the user through an LDA document theme generation model, wherein the theme probability distribution corresponding to each document is the documentTopic probability distribution for a user
Step 1-4, for the user social relationship data, obtaining the text information generated by the user friends according to the user social relationship, and obtaining the interest distribution of the user friend circle in the same way
Wherein: the above steps 1-2, 1-3 and 1-4 are carried out in parallel.
The second stage comprises the following specific implementation steps:
step 2-1, substituting the new sign-in data of the user into f (x) to obtain probability density, and calculating the logarithmic mean value of the probability density
Step 2-2, for the newly generated text data of the user, passing through a formula
The topic probability distribution of the new text can be calculated
Where n (k) is the number of words in the text belonging to topic k, αkAre the model parameters obtained during the LDA training process.
Step 2-3, extracting text data generated by new friends for friends newly handed by the user, and distributing interest to the new friends in the same way
Wherein: the above steps 2-1, 2-2 and 2-3 are carried out in parallel.
The third stage comprises the following specific implementation steps:
step 3-1, obtaining a probability density logarithmic mean threshold S capable of distinguishing authenticity of user identity through geographical position through trainingc(ii) a Tong (Chinese character of 'tong')Obtaining a threshold value D of Jensen-Shannon divergence of two probability distributions capable of distinguishing authenticity of user identity through text information by trainingTAnd DF。
Step 3-2, for the subject probability distribution of the user text historical data
And subject probability distribution of user newly generated text
JS divergence for calculating two probability distributions
Step 3-3, interest probability distribution of historical data of user social relationship
And interest probability distribution of new friends of user
JS divergence for calculating two probability distributions
Step 3-4, for all mobile internet users, according to all characteristic values obtained by calculation in steps 3-1, 3-2 and 3-3, respectively according to corresponding judgment standards S from three dimensions of geographic position, text information and social relationc、DTAnd DFGiving a judgment U of the user's identityC，UT，UFFusing the judgment of the authenticity of the user identity by the three dimensions to obtain a final judgment value U of the user identityI＝g(UC，UT，UF)。
Algorithm
User identity false identification algorithm based on behavior modeling (see figure 2 for specific flow)
Inputting: user-generated check-in information, text information, and social relationships
And (3) outputting: truth value of user identity
(1) Calculating the characteristic distribution rule of the user historical data in three dimensions according to the user check-in geographical position information, the text content sent by the user online behavior space and the social relationship of the user in the whole social network
And
executing the step (2);
(2) for a newly generated behavior record of a user, it is evaluated in three dimensions: calculating a log mean of probability density values of newly signed data
Calculating a topic probability distribution for newly generated text
Calculating interest distribution for newly established social relationships
The method specifically comprises the following steps:
obtaining probability density function f (x) of sign-in position distribution by adopting a check density estimation method according to historical sign-in data of users and friends of the users, substituting the data of the new sign-in of the users into the f (x) to obtain probability density, and calculating the mean logarithmic value of the probability density function
The above-mentioned
Obtaining topic probability distribution of new text
Wherein n (k) is textNumber of words in subject k, αkModel parameters obtained in the LDA training process; the interest distribution of the new friend is obtained by extracting the text data generated by the new friend in the same way
Executing the step (3);
(3) in the check-in information dimension, to be calculated
The value of (A) and a probability density logarithmic mean threshold value S obtained by training and capable of distinguishing authenticity of user identity through geographic positioncComparing, if greater than threshold, returning UCOtherwise, return UCExecuting the step (4) as false;
(4) subject probability distribution for user historical data in the text information dimension
And subject probability distribution of user newly generated text
Jensen-Shannon divergence calculating two probability distributions
The value of the JS divergence is compared with a threshold value D of JS divergence between two probability distributions obtained through trainingTComparing, if less than threshold, returning UTOtherwise, return UTExecuting the step (5) as false;
(5) in the social relationship dimension, text information capable of expressing interests in friend information is extracted according to the social relationship of the user, and the probability distribution of the topics of the user friend historical data is calculated in the same way
And subject probability distribution of user newly generated text
And JS divergence
Its value is compared with a threshold value DFComparing, if less than threshold, returning UTOtherwise, return UTExecuting step (6) as false;
(6) the judgment U for the user identity truth value returned according to the steps (3), (4) and (5)C，UT，UFThe judgment of the authenticity of the user identity in three dimensions is combined to obtain the final judgment value U of the user identityI＝g(UC，UT，UF)。
Experiments prove that the method is superior to the previous research in both accuracy and calculation time.
1 points of innovation of this project
1. And establishing a normal behavior model of the user according to the historical behavior data of the user.
2. And a more accurate identity recognition method is obtained by utilizing the complementary effect of the multiple dimensional behaviors.
3. The prior identity identification mode is distinguished, the identity identification mode does not depend on hardware equipment, is used as an identity identification of a user according to the behavior characteristics of the user, and has high reliability.
And (3) annotating: the terminology used in the present invention and the prior art can be found in the following.
[1]Bao J，Zheng Y，Wilkie D，et al.A survey on recommendations in location-based social networks[J]ACM Transaction on Intelligent Systems and Technology，2013.
[2]David M.Blei，Andrew Y.Ng，Michael I.Jordan.Latent Dirichlet Allocation[J]//Journal of machine learning researcn，2003，993-1022
[3]Wang X，Mccallum A，Wei X.Topical N-Grams：Phrase And Topic Discovery，With An Application To Information Retrieval[C]//Data Mining，2007.ICDM 2007.Seventh IEEE International Conference on.2007：697-702
[4]Jie Bao，Yu Zheng，Mohamed F.Mokbel.Location-based and preference-aware recommendation using sparse geo-social networking data.[J]//International Conference on Advances in Geographic Information Systems，ACM，2012
[5]Lichman M，Smyth P.Modeling human location data with mixtures of kernel densities[C]//Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining.ACM，2014：35-44.
[6]Cho E，Myers S A，Leskovec J.Friendship and mobility：user movement in location-based social networks[C]//Proceedings of the 17th ACM SIGKDD international conference 0n Knowledge discovery and data mining.ACM，2011：1082-1090.
[7]Yang Song，Zheng Hu，xiaoming Leng.Friendship influence on mobile behavior of location based social network users.[J]//Journal of Communications and Networks，2015.
Claims (1)
1. A user identity false-identification algorithm based on behavior modeling is characterized by comprising the following processes:
inputting: user-generated check-in information, text information, and social relationships
And (3) outputting: truth value of user identity
(1) Calculating the characteristic distribution rule of the user historical data in three dimensions according to the user check-in geographical position information, the text content sent by the user online behavior space and the social relationship of the user in the whole social network
And
executing the step (2);
(2) for a newly generated behavior record of a user, it is evaluated in three dimensions: calculating a log mean of probability density values of newly signed data
Calculating a topic probability distribution for newly generated text
Calculating interest distribution for newly established social relationships
The method specifically comprises the following steps:
obtaining probability density function f (x) of sign-in position distribution by adopting a check density estimation method according to historical sign-in data of users and friends of the users, substituting the data of the new sign-in of the users into the f (x) to obtain probability density, and calculating the mean logarithmic value of the probability density function
The above-mentioned
Obtaining topic probability distribution of new text
Where n (k) is the number of words in the text belonging to topic k, αkModel parameters obtained in the LDA training process; the interest distribution of the new friend is obtained by extracting the text data generated by the new friend in the same way
Executing the step (3);
(3) in the check-in information dimension, to be calculated
The value of (A) and a probability density logarithmic mean threshold value S obtained by training and capable of distinguishing authenticity of user identity through geographic positioncComparing, if greater than threshold, returning UCOtherwise, return UCExecuting the step (4) as false;
(4) subject probability distribution for user historical data in the text information dimension
And subject probability distribution of user newly generated text
Jensen-Shannon divergence calculating two probability distributions
The value of the JS divergence is compared with a threshold value D of JS divergence between two probability distributions obtained through trainingTComparing, if less than threshold, returning UTOtherwise, return UTExecuting the step (5) as false;
(5) in the social relationship dimension, text information capable of expressing interests in friend information is extracted according to the social relationship of the user, and the probability distribution of the topics of the user friend historical data is calculated in the same way
And subject probability distribution of user newly generated text
And JS divergence
Its value is compared with a threshold value DFComparing, if less than threshold, returning UTOtherwise, return UTExecuting step (6) as false;
(6) the judgment U for the user identity truth value returned according to the steps (3), (4) and (5)C，UT，UFThe judgment of the authenticity of the user identity in three dimensions is combined to obtain the final judgment value U of the user identityI＝g(UC，UT，UF)。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810043919.7A CN108268762B (en) | 2018-01-17 | 2018-01-17 | Mobile social network user identity identification method based on behavior modeling |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810043919.7A CN108268762B (en) | 2018-01-17 | 2018-01-17 | Mobile social network user identity identification method based on behavior modeling |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108268762A CN108268762A (en) | 2018-07-10 |
| CN108268762B true CN108268762B (en) | 2021-04-30 |
Family
ID=62775799
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810043919.7A Active CN108268762B (en) | 2018-01-17 | 2018-01-17 | Mobile social network user identity identification method based on behavior modeling |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108268762B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109598110A (en) * | 2018-12-10 | 2019-04-09 | 北京羽扇智信息科技有限公司 | A kind of recognition methods of user identity and device |
| CN111475738B (en) * | 2020-05-22 | 2022-05-17 | 哈尔滨工程大学 | Heterogeneous social network location anchor link identification method based on meta-path |
| CN116260715B (en) * | 2023-05-09 | 2023-09-01 | 国品优选(北京)品牌管理有限公司 | Account safety early warning method, device, medium and computing equipment based on big data |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103853841A (en) * | 2014-03-19 | 2014-06-11 | 北京邮电大学 | Method for analyzing abnormal behavior of user in social networking site |
| CN106202488A (en) * | 2016-07-19 | 2016-12-07 | 西北工业大学 | Estimation user is to the method for physical event distance |
| CN106600052A (en) * | 2016-12-12 | 2017-04-26 | 西安交通大学 | User attribute and social network detection system based on space-time locus |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9077744B2 (en) * | 2013-03-06 | 2015-07-07 | Facebook, Inc. | Detection of lockstep behavior |
-
2018
- 2018-01-17 CN CN201810043919.7A patent/CN108268762B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103853841A (en) * | 2014-03-19 | 2014-06-11 | 北京邮电大学 | Method for analyzing abnormal behavior of user in social networking site |
| CN106202488A (en) * | 2016-07-19 | 2016-12-07 | 西北工业大学 | Estimation user is to the method for physical event distance |
| CN106600052A (en) * | 2016-12-12 | 2017-04-26 | 西安交通大学 | User attribute and social network detection system based on space-time locus |
Non-Patent Citations (1)
| Title |
|---|
| 基于位置社交网络的用户行为建模与研究;郭昊;《中国优秀硕士学位论文全文数据库(电子期刊)》;20180115;第2018年卷(第1期);I138-2181 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108268762A (en) | 2018-07-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zannettou et al. | On the origins of memes by means of fringe web communities | |
| Su et al. | De-anonymizing web browsing data with social networks | |
| García-Recuero | Discouraging abusive behavior in privacy-preserving online social networking applications | |
| Gervais et al. | Quantifying web-search privacy | |
| Liu et al. | Weighted graph clustering for community detection of large social networks | |
| CN106874253A (en) | Recognize the method and device of sensitive information | |
| CN108268762B (en) | Mobile social network user identity identification method based on behavior modeling | |
| CN107944032B (en) | Method and apparatus for generating information | |
| Zulfadhilah et al. | Cyber profiling using log analysis and k-means clustering | |
| CN104636408A (en) | News authentication early warning method and system based on user generated content | |
| US11036818B2 (en) | Method and system for detecting graph based event in social networks | |
| Wang et al. | Identity theft detection in mobile social networks using behavioral semantics | |
| Khayyambashi et al. | An approach for detecting profile cloning in online social networks | |
| CN104166726A (en) | Microblog text stream oriented sudden keyword detecting method | |
| Nouh et al. | Towards designing a multipurpose cybercrime intelligence framework | |
| Han et al. | Linking social network accounts by modeling user spatiotemporal habits | |
| Wang et al. | Fusing behavioral projection models for identity theft detection in online social networks | |
| Bródka et al. | Profile cloning detection in social networks | |
| Cheung et al. | Evaluating the privacy risk of user-shared images | |
| Figueira et al. | Current State of the Art to Detect Fake News in Social Media: Global Trendings and Next Challenges. | |
| Zhou et al. | An efficient victim prediction for Sybil detection in online social network | |
| Singh et al. | Fake (sybil) account detection using machine learning | |
| Dewan et al. | Multiosn: realtime monitoring of real world events on multiple online social media | |
| Shi et al. | SRTM: A Sparse RNN-Topic Model for Discovering Bursty Topics in Big Data of Social Networks. | |
| Al-Khateeb et al. | Developing a conceptual framework for modeling deviant cyber flash mob: a socio-computational approach leveraging hypergraph constructs |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |