Research On Equilibrium Optimization Mechanisms For Security And Utility Of Provenance Sanitization

In the information age,thousands of data objects are generated every minute.These data objects are vital to the decision-making of various organizations and institutions.However,sources of these data are diverse,and people should verify data provenance to validate the quality and credibility of these data objects before using them.Data provenance records the historical information of data objects,which is usually represented as a directed acyclic graph,referred to as the provenance graph.The provenance graph may contain a variety of sensitive information and the leakage of sensitive information may cause unexpected consequences.As a result,the technology of provenance sanitization come into being.To obtain a secure provenance graph,provenance sanitization redacts nodes,edges or indirect dependencies without violating the constraints of the provenance graph.There still exists two common drawbacks in existing provenance sanitization mechanisms.The first one is the low generality.Specifically,one sanitization mechanism can only sanitize one specific type of sensitive elements and can not be used to deal with comprehensive sanitization requirements of users at the same time.The second one is low customizability.Specifically,existing provenance sanitization mechanisms cannot adjust or optimize security and utility according to the preferences of different users.To this end,this thesis proposed a primitive-based general framework of provenance sanitization.The framework can deal with comprehensive sanitization requirements.And on this basis,an approach is proposed to enable the equilibrium optimization for security and utility of provenance sanitization and then applied in domains of supply chain and scientific workflow.This thesis mainly includes the following three contents.(1)A primitive-based general framework of provenance sanitization is proposed to deal with comprehensive sanitization requirements.Firstly,we introduce the types of sensitive elements and constraints involved in provenance sanitization.Secondly,we thoroughly analyze existing sanitization mechanisms and formally define the provenance sanitization primitives.We also divide the general provenance sanitization process into three stages of sanitization,repairing and constraint verification.On the basis of that,we propose a method for constructing the space of possible sanitization strategies by selecting and composing sanitization primitives stage by stage.Finally,a primitive-based general provenance sanitization algorithm is designed and implemented.Its feasibility is then verified on the public datasets.(2)An approach is proposed to enable the equilibrium optimization for security and utility of provenance sanitization.Firstly,the basic idea of using the hierarchical sorting method to balance the utility and security of the sanitized graphs is introduced.Secondly,in view of the large time and space problems of the sanitization strategy space constructed by the general sanitization framework,the key factors affecting utility of the sanitized graphs are theoretically proved or experimentally validated.So the sanitization strategy space can be pruned to keep only sanitization strategies with provenance utility higher than given threshold.Finally,an equilibrium optimization algorithm for security and utility of provenance sanitization is implemented and verified on the public provenance dataset.The experimental results show that equilibrium optimization algorithm can generate similar sanitization views with a much higher efficiency in comparison with the primitive-based general sanitization algorithm without strategy pruning.(3)We analyze the domain characteristics of the supply chain and scientific workflow,and apply the proposed general framework and the equilibrium optimization algorithm to these domains.F irstly,we introduce the application of the equilibrium optimization mechanism for provenance sanitization in the supply chain field to illustrate the generality of the algorithm.Secondly,in order to illustrate the customizability of equilibrium optimization for provenance sanitization,we analyze the domain characteristics of the scientific workflow field,and apply the equilibrium optimization mechanism for provenance sanitization in the field of scientific workflow with domain constraints.Finally,a specific experiment is designed in the corresponding provenance graph example to illustrate the feasibility and customizability of the algorithm in the application field.