Privacy-preserving and efficient outsourcing computation scheme for principal component analysis

Abstract Currently, the rapid development of Internet of Things (IoT) technology and the deepening digital transformation of traditional industries have driven substantial growth in the Machine Learning as a Service (MLaaS) market to meet massive data processing demands. Principal Component Analysis (PCA), as a widely used dimensionality reduction algorithm in fields such as image processing and finance, involves multiple matrix operations with high computational complexity. As a result, clients with low-capability devices often choose to outsource these computations to cloud servers. However, the privacy and security issues during data outsourcing have become critical challenges that need to be addressed urgently. In this paper, we propose a secure outsourcing computation scheme for PCA based on Householder matrix transformation and Double Random Perturbation Encryption (DRPE) techniques, aiming to protect the privacy of user inputs and outputs. The proposed scheme not only significantly reduces the computational complexity on the client side but also enables the client to detect malicious behaviors from the cloud server with a non-negligible probability, thereby enhancing the security of the data during the outsourcing process and the reliability of the computation results. Furthermore, the proposed scheme adopts a non-interactive design, effectively reducing communication overhead. Through comprehensive security analysis and performance evaluation, we demonstrate the superiority of the proposed scheme in terms of privacy protection, computational efficiency, and verification capability, providing a novel solution for secure outsourcing computing in IoT environments.