Secure genomic data evaluation in cloud environments

Abstract

Security in clouds often focuses on preventing unauthorized access to confidential information. However, cloud providers might also be a source for loss of confidentiality and are generally considered in risk models as honest but curious (HBC). The wide availability and high volume of genomic data improves advances in biomedical research, but outsourcing genomic data processing to cloud providers presents new challenges and risks due to the confidentiality of such data and the critical consequences of a possible loss. We present techniques to securely delegate Genome-Wide Association Study (GWAS) data into clouds using encrypted data. The protocol is based on homomorphic properties of well known encryption algorithms. The protocol can also be used to amend existing applications by software patches of binaries. In the present paper we introduce some practical extensions to our algorithm to improve its efficiency. Additionally we extend the algorithm to support novel optimizations, including Single Operation Multiple Data (SIMD) while preserving its homomorphic properties. We evaluate the protocol by a proof-of-concept implementation of minor allele frequency and chi-squared statistics computations on real-life genomic data to investigate practicability, and discuss variants and extensions to increase the prototype's efficiency.