Secure Computation of Top-K Eigenvectors for Shared Matrices in the Cloud

TL;DR

This paper presents a secure, scalable method for computing the top-k eigenvectors of shared matrices in the cloud, ensuring data confidentiality using encryption and perturbation techniques with low client costs.

Contribution

It introduces a novel secure iterative algorithm for top-k eigenvector computation that preserves data privacy in cloud environments.

Findings

Method is scalable to large matrices

Ensures data confidentiality with encryption and perturbation

Requires low client-side computational costs

Abstract

With the development of sensor network, mobile computing, and web applications, data are now collected from many distributed sources to form big datasets. Such datasets can be hosted in the cloud to achieve economical processing. However, these data might be highly sensitive requiring secure storage and processing. We envision a cloud-based data storage and processing framework that enables users to economically and securely share and handle big datasets. Under this framework, we study the matrix-based data mining algorithms with a focus on the secure top-k eigenvector algorithm. Our approach uses an iterative processing model in which the authorized user interacts with the cloud to achieve the result. In this process, both the source matrix and the intermediate results keep confidential and the client-side incurs low costs. The security of this approach is guaranteed by using Paillier Encryption and a random perturbation technique. We carefully analyze its security under a cloud-specific threat model. Our experimental results show that the proposed method is scalable to big matrices while requiring low client-side costs.