Anonymization-based Research On Privacy Preserving Data Publishing In ERP Systems

With the rapid development of information technology such as the Internet,all areas are rich in the information data.The ability of data collecting,data analyzing and data mining has been greatly improved in all fields,especially,data mining techniques can help explore the tremendous value of data.Data release is an essential step of data mining.It can take advantage of the third-party technology and public wisdom to brainstorm in order to make full use of the data value and provide strategic decision-making better.However,privacy leaks and information security occur when publishing the data.This problem has become a bottleneck for the further development of data analysis and data mining.In order to ensure privacy when releasing the data,the usual method is to use meaningless symbols to replace a single or some of the individual attributes which can uniquely identify the individual,but this approach cannot take good effect.The attacker can also identify the user through background knowledge or other information,and then get access to the user's sensitive information.Academics have put forward lots of techniques and methods,of which anonymization technology is a classic privacy preserving method.Due to the high authenticity and good quality of its data,ERP information system owns a high value in data publishing and analyzing.The paper takes ERP information system as the background and mainly studies the privacy preserving method on data publishing in this system.The main work and contribution can be described as follows:First,the paper proposes a privacy preserving method based on k-anonymization according to the attack model of ERP information system.First of all,the paper analyzes the experimental data set SAP GBI 2.3 and constructs the attack model based on sales order after proposing the appropriate data structure and related assumptions with the consideration of the general characteristics of data in ERP information system.The paper also introduces a practical data utility metric.Furthermore,the paper develops a Weighted Matching K-anonymization Algorithm(WMKA)according to the proposed attack model,and proves the validity and superiority of the algorithm by comparing with the other two algorithms according to the data utility metric.The great contribution of this part is proposing a general attack model according to the characteristics of data in ERP information system and developing an effective anonymization algorithm.Second,the paper proposes a new data structure and anonymization methods for a specific field of ERP information system-railway ERP system.Due to the diversity of data in railway ERP system,the paper puts forward a GeoSocial Network(GSN)model based on hypergraph with the consideration of social network information and geographical location information,and constructs the attack model and anonymous model based on the GSN model.The paper also defines several data utility metrics.Furthermore,the paper develops the(k,m)-anonymization algorithm and(k,m,l)-anonymization algorithm of GSN according to the above models.and evaluates the experimental results in different periods through a large number of experiments with the measure of data utility metrics in order to prove the validity of the algorithm.The great contribution of this part is proposing a GSN model based on hypergraph according to the complex characteristics of data in railway ERP system data.The paper also puts forward a practical attack model and anonymization model and develops a better set of anonymization algorithm after defining reliable data utility metrics.