Deploying Scalable and Secure Secret Sharing with GPU Many-Core Architecture

Abstract

Secret sharing is an excellent alternative to the traditional cryptographic algorithms due to its unkeyed encryption/decryption and fault tolerance features. Key management hassle faced in most encryption strategies is removed from users and the loss of a certain number of data copies can be tolerated. However, secret sharing schemes have to deal with two contradictory design goals: security and performance. Without keys' involvement, large security margin is expected for the illusion of being computationally secure. In the meantime, such design will degrade the performance of encrypting and decrypting secrets. Thus, secret sharing is mainly for small data such as keys and passwords. In order to apply secret sharing to large data sets, this paper redesigned the original schemes to balance the security and performance. With sufficient security margin, Graphics Processing Unit (GPU) is adopted to provide the performance satisfaction. The proposed secret sharing scheme with GPU acceleration is a practical choice for large volume data security. It is particularly good for long-term storage for its unkeyed encryption and fault tolerance. Performance analysis and experimental results have demonstrated the effectiveness and efficiency of the proposed scheme.