Object detection framework to generate secret shares

• Published in: Multimedia tools and applications 2020-09, Vol.79 (33-34), p.24685-24706
• Main Authors: Agarwal, Ayushi, Deshmukh, Maroti, Singh, Maheep
• Format: Article
• Language: English
• Subjects: Algorithms; Computer Communication Networks; Computer Science; Correlation analysis; Data Structures and Information Theory; Encryption; Image reconstruction; Image transmission; Multimedia Information Systems; Object recognition; Special Purpose and Application-Based Systems
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-020-09169-x

Nowadays, the sharing of images via the Internet has been widely used. The security concern during transmission of images has been a very important issue, as images contain a lot of crucial information. To prevent access to this information by any unauthorized user various encryption schemes have been developed. An image may consist of a large number of pixels and the encryption of the whole image takes more time. To overcome this problem, we proposed an algorithm that focuses on encryption of only those objects which contain the major information of an image instead of encrypting the complete image, this saves the time required in encryption as only the objects contained in the image are encrypted. The proposed algorithm exploits the use of object detection and image secret sharing. Object detection is done using the “You Only Look Once (YOLO)” algorithm. Further, the objects detected are encrypted using (n,n) modular arithmetic secret sharing scheme. The quantitative measures like correlation, SSIM, RMSE, PSNR has been used to evaluate the performance of the proposed algorithm on COCO dataset. The experimental results show that the proposed algorithm is lossless i.e. original and reconstructed images are exactly the same. The proposed algorithm is efficient and can be used in a broad range of applications.