Privacy-preserving multi-source content-based image retrieval for IoT in cloud computing

As Internet of Things (IoT) devices develop, an increasing number of images appear in our everyday lives. The emergence of massive amounts of image data has increased the burden on local IoT devices. As IoT owners try to outsource to the cloud server, privacy disclosure may occur during the outsourcing process. In this study, we designed a secure image-retrieval scheme that supports multiple sources. In our solution, we first encrypt the image features by improving a secure multiparty sum protocol, which can effectively support the similarity calculation of encrypted features gathered from multiple sources. Then, IoT owners store the feature vectors locally to avoid disclosing similarity information to the cloud. Finally, we created an index to improve search efficiency by locality-sensitive hashing. Security is analyzed to demonstrate that the privacy of images and matching results have been well protected against the cloud server. In addition, the performance analysis and experimental results show that the proposed scheme achieves highly competitive results on real datasets.