Development of Security Aware Content Based Image Retrieval System using Lightweight Trapdoor Verification in Cloud Environment

Abstract

Demand for personal privacy protection, safe cloud storage, and search over encrypted information have all grown to be critical issues due to the rapid development of cloud services. The movement towards safe computation has received a lot of attention, particularly asymmetric scalar-product-preserving encryption (ASPE) and homomorphic encryption (HE). Although ASPE has the capacity to effectively encrypt and compare cipher texts, its reliance on the assumption that users can be completely trusted in the actual world and potential key leakage issues make it an impractical technique. In this research develop a content-based image retrieval system model that is security conscious and utilize the lightweight trapdoor model for verification, initially the input are collected from the standard repository and saved in an encrypted format. The image is encrypted using optimal chaotic map-based encryption, and the keys are produced on the user side to provide the image privacy and security. The image obtained after encryption is called as ciphered image. Hence for the purpose of image retrieval, the ciphered image is kept in the cloud. The relevant ciphered images are provided to the user whenever a search request is made by the query user using the trapdoors. Using an efficient hybrid distance-based clustering method, the cloud's most pertinent ciphered images are chosen, and the user who requests them is given access to the relevant image clusters. The proposed hedge learner optimization is used to optimize the hybrid distance based clustering of images. Finally the cloud image that the user currently has been ciphered is decrypted using the owner's key.Based on the achievements at TP 80, the HLO Optimized Chaotic Encryption model achieve the MSE, PSNR, RMSE and SSIM values of 0.04, 61.67, 0.21 and 93.97 respectively. Similarly HLO Optimized Hybrid Distance based Clustering model achieves the F1 measure, precision and recall values during NOR 500 is 94.73%, 95.20% and 95.60% respectively.