The Construction Of Early-warning Intelligence Model Against Phishing Scam

With the concept of the "Internet Plus" being put forward,the development of Internet business models and the spread of mobile payments into all aspects of social life,the behavior of payment becomes the core information dimension in the era of big data.As a negative product of the Internet,the phishing scam further spreads in the world with the development of mobile payments.It steals account funds by means of camouflaging trusted entities,creating malicious links and so on,which seriously jeopardizes the security of accounts of network users.The traditional intelligence work is facing challenges from new cybercrimes,and it is an irresistible trend to conduct "Internet plus" public security intelligence work and combine the artificial intelligence with fighting against crimes.After expounding the research background of the phishing scam,the paper studies and analyzes the specific connotation and means of phishing scams as well as various challenges facing the early-warning intelligence.Combine with the Internet Plus concept,it constructs the module framework of early-warning intelligence model against phishing scam which based on data mining method.The module framework,with early warning intelligence analysis as the core part,includes the early-warning information framework of different modules such as source,target,collection and processing,decision-making and so on.The paper concentrates on how to implement early-warning intelligence analysis through the construction of early-warning intelligence model.After studying the phishing scam detection technologies and related theoretical research results at domestic and overseas,and referring to the authentication theories and the phishing scam process,the author proposes two early-warning intelligence models based on data mining methods,according to the characteristics that abnormal operations and precursory evidence may inevitably occur due to hijacked online payment accounts after the occurrence of phishing scams.The first one is the early-warning intelligence model against phishing scams based on semi-supervised anomaly detection,where the author extracts typical characteristic variables of ordinary accounts and constructs a model through multivariate Gaussian distribution with observation of previous phishing scams.Then,the author obtains the optimal threshold through the F1-score evaluation model and substitutes it into the test group to validate the model performance in order to prove its practical significance in the early-warning of phishing scams.The other one is an early-warning intelligence model against phishing scams based on Bayesian analysis.The author constructs the model through the hypothesis of phishing scams,initial probability and case evidence,estimates the probability of occurrence of phishing scams by establishing the likelihood ratio and calculating posterior probability and continuously updates the evidence list to realize the real-time update of probabilities.At last,the author verifies the model’s accuracy with the May 15 Phishing Scam in L city as an example.