Research On Key Technologies Of Secret Image Sharing For Visual Internet Of Things

The rapid development of the information age makes rich multimedia information transmitted in the network,and images are widely used as an effective information carrier in multimedia.At the same time,more focus on image and other multimedia information acquisition and processing of visual Internet of Things(Io T)technology emerged,along with its technical development of security issues are also drawing attention,people’s awareness and demand for the protection of privacy information in image data is increasingly strong,so how to effectively ensure the security of image information in the visual Internet of Things has become an important issue in the information age.Secret image sharing technology has become a hot research topic because it can guarantee the integrity of secret image information while ensuring the confidentiality of image information using the idea of distributed technology.Currently,the application of secret image sharing technology in visual Io T systems is widely researched,however,there are some problems in its application.First,in the field of smart transportation in visual Io T,there is a lack of schemes that take into account the distributed storage of images with multi-level access to the contents of different regions of images by authorized parties while embedding additional data into images to avoid additional data access;in traffic monitoring scenarios,the visual quality of important content regions of recovered images after the shadow images are destroyed is poor,and the robust performance of sharing schemes needs to be improved urgently.Based on the above,two schemes in this thesis are proposed to enhance the application prospects of secret image sharing in visual Io T,mainly as follows.1.Propose a hierarchical secret sharing scheme for secure access to smart traffic images.First,a secret image sharing algorithm based on reversible data hiding is proposed.The algorithm uses the correlation between image pixels to retain certain differential information in terms of image pixel blocks,which is combined with the image sharing operation process to realize large-capacity reversible data embedding.Then,the algorithm is extended to realize hierarchical sharing of images and reversible data.The deep learning algorithm is used to divide different areas of traffic images in the intelligent transportation application scenario according to the importance of the content,and to carry out multi-level sharing while embedding the same level of secret data accordingly using the above algorithm and sending the shared share to different levels of authorized parties for storage.When recovering the secret images and data,only the set of authorized parties corresponding to the rank is reconstructed by Lagrange interpolation formula for the same rank image and extracts the secret data.It is proved through experiments that the scheme has high data embedding capacity,the minimum average embedding rate can reach3.63 bpp,and the secret images and data can be lossless reconstructed.2.Propose a verifiable and recoverable robust secret sharing scheme for traffic surveillance images.The scheme takes advantage of the property that Permutation Ordered Binary(POB)digital system can compress binary sequences and lossless reconstruct them,and embeds the relevant information into the secret image to compress it to achieve the robustness of the sharing scheme.First,the scheme embeds the image recovery information generated using image pixel correlation into the mapping of the secret image,and uses these data to generate verification information together with the embedding.Then the POB digital system algorithm is applied to compress the image bits and perform the second embedding of the verification information,and the shared image generates the shadow image.The inverse process of the POB algorithm is performed at the receiver side during recovery to extract the double verification information for two-level verification,locate the tampered region of the image,and then reconstruct the tampered region using the embedded recovery information.It is demonstrated experimentally that the proposed scheme has certain robustness in large-area shadow image destruction,the visual quality of reconstructed important content regions in traffic surveillance images is better,and the average PSNR of reconstructed images can reach up to 51.12 dB.