Practical and Secure Outsourcing of Matrix Computations to the Cloud

Cloud computing enables a cost effective outsourcing of storage and resource-intensive computations. Secure outsourcing of data and computation is challenging in this emerging computing model. In particular, outsourcing of sensitive computations should be assured in terms of input privacy and cheating detection. Existing approaches use either expensive homomorphic encryption which is not yet efficient enough for practical applications or secure tamper-proof hardware that are expensive and not scalable. Works on devising a real-world framework enabling such secured outsourcing are still needed. In this paper, we propose practical protocols for secure outsourcing of matrix algebra using randomization and without using cryptography. We address the issues of a real deployment in terms of distributed, reliable and secure storage, secure data transfer and computation on multiple, non-colluding clouds. We present the architecture and the APIs of our framework as well as some experimental results demonstrating the effectiveness of the proposed approach.