Secure Computing through Homomorphic Encryption on a Peer-Servicing Public Cloud Computing Platform

摘要

Cloud computing is able to redistribute the computing resources, and then reduces the maintenance cost, thereby increases profits. Ensuring data security and maintaining confidentiality of personal information are important issues of cloud computing security. In order to avoid malicious attacks and data theft in the applications of cloud computing, homomorphic encryption provides a novel technique for computing with zero-knowledge privacy. Fully homomorphic encryption, proposed by Gentry in 2009, allowed the execution of all kinds of secured computation without the secret key. Data privacy in the public cloud could be achieved in this way. This study proposes a cloud-computing framework utilizing fully homomorphic computation for processing the customer's pre-encrypted data. The proposed framework is able to compute and to verify the result without the need of decrypting the data in the whole process. The integration of homomorphic scheme with garbled circuits provides a mechanism of verifiable computation. It avoids the security vulnerability where data is kept by an entrusted third party. In addition, this paper builds a system on a peer-servicing public cloud platform powered by the P2P Hadoop. Accordingly, there is no single point of failure in this high-availability system. The experiments show that the security of this system meets the standards. In conclusion, the proposed framework has confidentiality, integrity and availability complying with core criterions of information security.