A Survey Paper on Secure Privacy Preserving Structure for Content Based Information Retrieval on Large Scale

Abstract

It is very essential to protect personal confidential data that we share or search through web. Previously there are number of privacy preserving mechanism has been developed. Here we develop a new privacy protection framework for huge- content-based information retrieval. We are offering protection in two layers. Initially, robust hash values are taken as queries to avoid revealing of unique features or content. Then, the client has to select to skip some of the bits in a hash value for increasing the confusion for the server. Since we are reducing information it is not so easy for servers to know about interest of the client. The server needs to give back the hash values of all promising candidates to the client. The client will find the best match by searching in the candidate list. Because we are only sharing hash values between server and client the privacy of client and server will be protected. We begin the idea of tunable privacy, where we can adjust level of privacy protection according to the policy. We can realized it by hash based. It can be realized through piecewise inverted indexing based on hash. We have to divide extracted feature vector into pieces and index each and every piece with a value. Every value is linked with an inverted index list. The framework has been comprehensively tested with very huge image database. We have estimated both privacy-preserving performance and retrieval performance for those content recognition application. Couple of robust hash algorithm is being used. One is based on discrete wavelet transform; the other is based on the random projections. Both of these algorithms demonstrate acceptable recital in association with state-of-the-art retrieval schemes. We believe the bulk voting attack for guesstimate the query recognition and sort. Experiment results confirm that this attack is a peril when there are near-duplicates, but the success rate is depends upon the number of distinct item and omitted bits, success rate decrees when omitted bits are increased.