A Study on Secure Outsourcing of Large Scale Computations to Cloud

Abstract

At present, cloud computing has become the most prominent field for the clients which make them relax to outsource their expensive computations. Based on a pay-per-use model, a client without enough computational power can easily outsource large scale computational tasks to a cloud. But this relaxation also brings some anxious like as data confidentiality, storage overhead and trustworthiness of cloud and so on. To relief from these threats, a perfect cloud outsourcing protocol is needed. A cloud outsourcing system is perfect if it can minimize the client’s overhead as well as it has trust. So, the problem is how a client can send his overload task to cloud by privacy preserving and verifiable way. Since mathematical computations always have contribution on various scientific and engineering tasks, we are motivated to design a secure, efficient and verifiable cloud outsourcing protocol for some large-scale mathematical computations such as lyapunov equation, linear regression. Lyapunov equation needs for stability analysis. It is applied to the power system analysis also. To solve it by cloud, we use affine transformation that actually transfers the problem linearly. These transformation computes in a such way that from transferred result original result can be found. Linear regression is studied for with constrained and without constrained. Unconstrained linear regression that is studied here in two ways. First of all, it is studied by hiding its dimension and then transferred the problem by random permutation. Because of increasing dimension, efficiency of this method is decreasing but increased security. It is the trade off between efficiency and security. Second, for transforming it, we use the idea of chaotic map and frobenius matrix. Chaotic map is one of the most used random number generation algorithms and it has some advantages using frobenius matrix over diagonal matrix. At last, constrained linear regression outsourcing is done by affine mapping like as lyapunov equation. Actually, We proposed protocols for outsourcing these problems and studied these protocols in a privacy preserving, efficient and verifiable way. Real cloud is also invoked and a comparison between real and simulation cloud is also showed. Theoretical and experimental results confirm the effectiveness and efficiency of our protocols.