PVIDM: Privacy-preserving verifiable shape context based image denoising and matching with efficient outsourcing in the malicious setting

With the emerging techniques of wireless communication and cloud computing, large volumes of multimedia data are outsourced from resource-constrained users to the cloud with abundant resource for both delegated storage and computation. Unfortunately, there exists a risk of users’ image privacy leakage in the process of outsourcing to the cloud in the malicious setting. Most of the existing work achieved privacy-preserving verifiable image feature extraction and matching by using public key (fully) homomorphic encryption (FHE), and the heavy computational overhead and communication overhead cannot adapt to resource-constrained mobile devices. Other works disabled to achieve image denoising in the encrypted domain or only focused on the scale-invariant feature transform (SIFT) descriptor that is inappropriate for position-sensitive feature extraction. To address these issues, in this paper, a privacy-preserving verifiable shape context based image denoising and matching protocol PVIDM with efficient outsourcing in the malicious setting is proposed. To achieve this end, by exploiting any one-way trapdoor permutation, an efficient privacy-preserving verifiable image denoising scheme PPVID is firstly proposed to improve the accuracy of image matching, without exploiting public key FHE. Then, based on PPVID, a lightweight shape context based privacy-preserving verifiable image feature extraction and matching protocol PPVIM is devised, where secure and efficient comparison and counting protocols in the encrypted domain are respectively proposed. Both image privacy and image matching result privacy are well protected, and the correctness of image matching result can be efficiently verified. Finally, formal security proof and extensive simulations on real-world data sets demonstrate the efficiency and practicability of our proposed PVIDM. Especially, both the computational complexity for the usage of any one-way trapdoor permutation at data owner's/user's end and the size of the authentication tag generated by the cloud are O(1), independent to both the number of database images and the number of pixels in each image.