AS5: A Secure Searchable Secret Sharing Scheme for Privacy Preserving Database Outsourcing

Abstract

Abstract.Researchers have been studying security challenges of database outsourcing for almost a decade. Privacy of outsourced data is one of the main challenges when the "Database As a Service" model is adopted in the service oriented trend of the cloud computing paradigm. This is due to the insecurity of the network environment or even the untrustworthiness of the service providers. This paper proposes a method to preserve privacy of outsourced data based on Shamir's secret sharing scheme. We split attribute values into several parts and distribute them among untrusted servers. The problem of using secret sharing in data outsourcing scenario is how to search efficiently within the randomly generated pool of shares. In this paper, at first, we customize Shamir's scheme to have A Searchable Secret Sharing Scheme (AS4) that enables the efficient execution of different kinds of queries over distributed shares. Then, we extend our method for sharing values to A Secure Searchable Secret Sharing Scheme (AS5) to tolerate statistical attacks based on adversary's knowledge about outsourced data distribution. In AS5 data shares are generated uniformly across a domain to prevent information leakage about the outsourced data.

Enforcing Privacy in Cloud Databases

“…Finally, up to n − k participants may disappear without compromising data availability; and message authentication code or signature can be applied to guarantee data integrity. Thus, secret sharing is a promising solution to security challenges in cloud data outsourcing [35] since one can easily imagine participants being CSPs. In the following subsections, we review the secret sharing-based approaches that aim at outsourcing databases, and then discuss them globally.…”

“…Moreover, intrinsically sensitive attributes such as social security numbers are stored locally at the user's. Eventually, fragmentation-based approaches yield little overhead on query computation, but are vulnerable in cases of CSP collusion and CSP inference on data updates [35]. Additionally, retaining sensitive data at the user's requires local storage capacities, which is incompatible with our data outsourcing scenario.…”

“…Statistical attacks refer to an adversary obtaining some information about ciphertexts, i.e., encrypted data, through prior knowledge about plaintexts, i.e., clear data [35]. For instance, the adversary may know plaintext distribution or frequency.…”

“…Original data are divided using order-preserving polynomials such that the order of shares is the same as that of original data. However, while this solution allows efficiently processing any kind of queries, including updates, it is susceptible to statistical attacks [35].…”

“…Moreover, it provides stronger security than Agrawal et al's scheme. It is indeed secure against frequency attacks, and an extension uses distribution perturbation to improve its robustness against statistical attacks in general [35]; but at the price of a higher computational overhead. Table 1 provides a comparison of secret sharing-based database outsourcing methods with respect to the queries that can run directly on shares and security features beyond privacy and availability, i.e., integrity checks and robustness against statistical attacks.…”