Access Policy For Attribute-Based Encryption

Cloud storage is a novel architecture of data storage. It brings larger developments of information technology industry. A lot of companies offer cloud storage services for customers, such as Simple Storage Service (3S) of Amazon, Openstack of RackSpace, Azure of Microsoft and so on. The core of cloud storage are storage devices, customers can access these devices via Internet and acquire various data servers.A crucial challenge in cloud storage is the security of outsourced data. It is important for service providers to ensure privacy and security of such data. Most of researchers have paid attention of data security of cloud. Attribute-based encryption (ABE) is a novel cryptographic primitive which can provide ciphertext access control for cloud. There are four advantages of ABE based access control mechanism:1) it is fine-grained access control mechanism; 2) it can implement one-to-many encryption; 3) data access control is maintained by the customer instead of the service provider; 4) the security properties of data are derived from cryptography.The access privileges are expressed by access policies in ABE and the performance of ABE schemes depends significantly on their access policy. There are a lot of forms of access policy, such as monotonous Boolean formula, access tree, linear secret sharing scheme, AND access structure and so on. In different scenarios, there are different requirements for different forms of access policy. This dissertation provides several optimization schemes for access policy of ABE. The contributions of this dissertation are included as follows:1) An anonymous policy-matrix for ABE is proposed. The attribute set of such scheme is mapped into a vector space as an orthogonal basis. The access policy can be viewed as a non-monotonic Boolean expression based on attribute set. The policy-matrix can be generated in three steps:a) the access policy is transformed into a disjunctive normal form; b) each conjunction expression of the disjunctive normal form is expressed by a linear combination of attribute vectors named conjunction vector; c) the policy matrix is consist of such conjunction vectors. The access policy can not be analyzed from the policy-matrix while the orthogonal basis is unknown. At the same time, we provide the interface of proxy encryption and decryption for to outsource computation and reduce cost for clients.2) A scalable access policy named block linear secret sharing scheme matrix (block LSSS matrix) is proposed. Block LSSS matrix is combined by a set of node matrixes. It is a sparse block matrix and each block of the matrix is mutual independent with others. The access policy is manageable cause of its strong scalability. In this paper, we propose an ABE scheme with block LSSS matrix. In such scheme, there are three policy managing methods:updating privilege, agency privilege and temporary privilege.3) A policy compression scheme based on access tree is proposed. The size of ciphertext is depended on the scale of access policy in ABE. The complex access policy incurs larger storage redundancy and computation cost. The dissertation proposes a method to adjustment the value of access tree (the logic structure is keep) and compress the node of access tree. In such way, the ciphertext size and computation cost are all reduced. However, the policy compression problem is NP complete and the optimal compression scheme can not be obtained in polynomial time. A greedy algorithm is proposed to obtain approximate compression scheme. Furthermore, the access policy can be compressed in "sub-tree" for the policy set with high overlap factor to increase compressing ratio.