Research On Recommendation Algorithm For Access Policy Data Sharing System

With the development of the Internet,data sharing has become a trend,and people pay attention to the security issues in the data sharing system.Access control technology can prevent illegal access by users.It is one of the important methods to protect data security.The attribute-based access control mechanism determines whether a user can access a specific data according to the attributes of the user and the data.It realizes the fine-grained access control and makes the large-scale data sharing system more flexible,more efficient,and easier to manage.However,attribute-based access control mechanisms require the data owner to assign a corresponding access policy for each shared data.With the development of large-scale data sharing systems,such as cloud storage,there are lot of number of users,data and attributes in the system,which increases the complexity of access policy assignation and reduces the efficiency of access policy assignation.In order to improve the efficiency of access policy assignation and reduce the cost of access policy management,this paper conducts a in-depth research on access policy recommendation algorithms,and proposes several access policy recommendation algorithms.The main contribution of this paper as follow:1)This paper first proposes a linear access policy,which can not only describe fine-grained access rights by a linear function,but also flexibly adjust the security level by updating the security threshold of shared data.Furthermore,this paper proposes a linear policy recommendation algorithm based on matrix factorization.This algorithm learns a policy matrix and a security threshold vector from the access log,and recommends a linear access policy to shared data.It reduces the workload of access policy assignation and improves the efficiency of the data sharing system.2)Recommending access policy by the real data of user may lead to the privacy leakage.Therefore,this paper proposes an access policy recommendation algorithm based on the random response mechanism,which satisfies the local differential privacy.This algorithm uses the random response mechanism to encode and randomly perturb the access data of user,which protects the user’s private information.Furthermore,this algorithm also decodes the perturbation code according to the maximum a posteriori probability decoding criterion,which reduces the error between the perturbed data and the real data,and ensures the utility of the access policy recommendation.3)In order to satisfy the dynamic network environment requirements in the data sharing system,this paper propose a linear access policy with dynamic update.This access policy can effectively reduce the overhead of access policy assign,when a new data or user takes part in the data sharing system.Furthermore,this paper proposes a linear access policy with dynamic update’s recommendation algorithm based on the matrix factorization algorithm.This algorithm improves the scalability of the data sharing system and reduces the difficulty of assigning and managing access policies.