| With the development of social networks,the process of human social networking and virtualization has accelerated.More and more social behavior between individuals in real life has been migrated to cyberspace,resulting in a large number of accounts,traces,and relationships.Faced with massive social network data,one of the main tasks of public security organs at this stage is to discover anomalies or suspicious targets through fragmented information,and to use known clues to further explore unknown clues and find hidden criminal organizations.Currently,social network analysis technology has achieved a series of research results and engineering applications in community discovery,influence analysis,user recommendations,and other aspects.However,there is relatively little research and implementation of social network analysis technology in the domestic public security industry,mainly due to the following reasons:Firstly,with the enhancement of users’ personal privacy and security awareness,the user information that can be obtained through public social networks and channels is gradually decreasing.Especially in the investigation of criminal cases,the suspects are aware of anti-surveillance measures and deliberately hide their identity information to reduce their online records and traces.Secondly,foreign social network analysis methods are relatively mature.However,due to differences in user language,social habits,cultural backgrounds,and other reasons,it is difficult to meet the business needs of domestic public security work.This thesis focuses on social network analysis for public security business,considering the common issues of data missing,platform differences,and information hiding in real social network environments.And selects account information with better universality as the research basis to alleviate the difficulties of social network analysis caused by data cold start and model non-universality.Specifically,the thesis starts from three aspects of social network account nickname,location trajectory,and friend relationship,forming an investigation framework based on "point,line,and surface" analysis paths.Firstly,by using simple and intuitive account nicknames,suspicious homogenous accounts are screened out from massive data.Secondly,by analyzing the trajectory characteristics of users in different social modes,"accompanying" accounts with similar spatiotemporal features are matched.Finally,based on social association relationships,the target account groups and members importance are extended for further analysis.The main research work and innovative achievements are as follows:(1)To address the problem of the public security department’s difficulty in associating unknown accounts across social networks and verifying user identities,this paper analyzes the account nicknames of social platforms such as We Chat,Weibo,and Douyin,determines that domestic social users often adjust keyword positions,change word pronunciations,and change nickname length to adapt to platform rules,and proposes a user alignment algorithm applicable to domestic social users.This algorithm achieves consistency judgment and cross-platform account association retrieval for any two social accounts.Experiments show that in the scenario of nickname consistency judgment,the accuracy rate of algorithm is about 18.67% higher than other similar algorithms on average;in the crossplatform account association finding scenario,the accuracy of algorithm is close to80%,and the accuracy of with different values is always better than other comparative method.This improves the public security department’s ability to identify the identities of unknown social accounts when conducting cross-social platform identity tracing.(2)To address the problem of the public security agency’s difficulty in discovering "accompanying" accounts with similar spatiotemporal features as the target account through social networks,this paper analyzes the common discrete and continuous spatiotemporal data in different usage scenarios of social platforms and proposes a dynamic spatiotemporal window-based companion analysis method.This method adjusts the time window and spatial distance to find "accompanying" accounts with similar spatiotemporal features as the target account,which improves the public security department’s intelligence analysis ability to proactively discover hidden accounts and mine behavioral clues.(3)To address the problem of the public security agency’s difficulty in locating criminal gangs and discovering important members from large-scale social network populations,this paper analyzes the topological relationships between social network accounts and the time attributes of account nodes,proposes the algorithm applicable to static finite networks and the algorithm applicable to dynamic temporal networks,and realizes the discovery and importance ranking of two types of social network groups’ key individuals.The experiments show that the proposed method in this paper is used for other methods of its kind in terms of network propagation dynamics,robustness and consistency check.This methods can help the public security department to locate core members,discover potential related persons or key individuals in growth,significantly reducing manual investigation workload,alleviating the situation of insufficient police force for comprehensive deployment,and improving the comprehensive information mining ability in the big data environment.The social network analysis system formed by the above methods has been applied in the investigation and handling of cases by the public security agency,and has played an important role in similar work scenarios such as social Io T expansion analysis and spatiotemporal collision of epidemic-related individuals,effectively improving work efficiency and reducing police force input. |
PDF Full Text Request