A privacy-preserving hyper-spectral remote sensing image analysis framework based on matrix outsourcing computation

Abstract Hyper-spectral remote sensing image provides more detailed spectral information of ground targets, while the joint sparse coding-based clustering (JSCC) algorithm are proposed for accurate hyper-spectral image segmentation. However, since hyper-spectral image data processing represent one of the most time-consuming computing tasks, it necessitates the acceleration achieved by cloud outsourcing computation. Furthermore, data privacy and result availability issues in cloud computing should be considered in real-world scenarios. In this paper, we propose a novel privacy-preserving outsourcing scheme for matrix multiplication and matrix pseudo-inversion in the JSCC algorithm, to achieve efficient and secure hyper-spectral remote sensing image analysis. The core of our scheme is to present a novel matrix blinding method, which obscures the original matrices through element-wise computation and utilizes secret key index sets rather than subtly designed matrices, thereby further reducing the storage cost of encryption and decryption operations. Additionally, we correspondingly present a new verification method that computes and compares randomly selected elements in obscured matrices to guarantee the availability of returned results. Experimental results demonstrate that our scheme is 4.15–10.79% superior to the state-of-the-art matrix blinding based privacy-preserving outsourcing computation schemes on average and achieves an efficiency improvement of 73.49–80.45% compared to the original hyper-spectral remote sensing image analysis algorithm.