Research On Personalized Anonymity Privacy Protection Based On Sensitive Attribute Classification

With the development of the times and the advancement of science and technology,the information technology industry is changing with each passing day.Data dissemination and data mining as an important means of daily information management and database application can assist humans in making scientific decisions.With the Internet,data mining and data sharing technologies developing rapidly,people's ability to collect,analyze,process and apply data is getting stronger and stronger,but this also poses a threat to the privacy of data,which will inevitably increase the risk of personal privacy leakage.Therefore,how to prevent sensitive private data leakage during data publishing has become a research hotspot in the field of information security.In 1998,the k-data anonymization protection model was first proposed by Samarati et al.,followed by the proposition of 1 diversity anonymity,p-sensitive k-anonymity,(a,k)anonymity and personalized anonymity models.However,in these existing models,the distribution of sensitive attributes is often not fully considered,and the similarity attacks are not well defended.Therefore,based on the personalized anonymous model,further research was carried out to pro-vide corresponding levels of protection for different levels of sensitive attributes,and to avoid the presence of more sensitive attributes in the same equivalence class in order to protect sensitive private data.Based on the p-sensitive k-anonymity model proposed by the predecessors,starting from the perspective of personalized anonymous privacy protection,considering the diversity of sensitive attribute distribution,a personalized(p,?i,d)k-anonymous model based on sensitive attribute grading is proposed.Set two parameter values ?iand d respectively,?iis the constraint value set for the sensitive attribute of different sensitivity level.The attribute with higher sensitivity needs to be set to with smaller constraint value;d indicates the sensitivity within the equivalent class.The weighted hierarchical distance of the attribute requires that the larger d is the better.The larger the d is,the larger the difference of the sensitive attribute value in the equivalence class is,meaning the better the diversity of the sensitive attribute is,the safer the data is,meanwhile the greedy clustering algorithm is adopted to reduce the information loss and improve the quality of data usage.Finally,the model is verified by experimental simulation,the result of which shows that compared with the global generalized p-sensitive k-anonymity algorithm,this newly proposed algorithm can effectively resist against small information loss,have reasonable running time and realize personalized anonymous privacy protection for similarity attacks.