A Secure Search for Outsourced Image Collection Based Content-Based Image Retrieval

Abstract

Various research fields have shown significant interest in the real-world applications of image retrieval in recent years. Content-Based Image Retrieval (CBIR) has become a prevalent technique that is gradually being integrated into retrieval systems. However, images require more storage than text documents, and cloud computing is often used to outsource them. For sensitive images, like those used in medicine, they must be encrypted before being sent to a third party.This study proposes a novel classification and retrieval technique to search for related objects in encrypted images. The proposed framework relies on Convolutional Neural Networks (CNN) and LSTM networks, which unlock the potential of secure, content-based image retrieval and mass encoding. In this technique, the original images are first processed using a CNN neural network, and their features are extracted. Next, the features of the encoded images are extracted by training a new CNN neural network using the weights and activation functions of the previous neural network. The feature set is then divided into two groups for training and testing, with the feature training portion used to train the LSTM neural network. The proposed method outperforms the original article in all of the evaluated parameters, according to simulation findings using MATLAB software and the results was viwed in excel based on the generated numbers. This research offers a promising approach to secure content-based image retrieval and mass encoding, which could have significant implications for sensitive fields like medicine.