PRIVACY PRESERVING FRAMEWORK FOR LARGE SCALE CONTENT BASED INFORMATION RETRIEVAL

Abstract

In this Internet era, multimedia content is drastically increasing over the Internet and distributed over thousands of servers. In order to nd content over the large scale database, content-based search strategies(CBIR) have been implemented. Recently, issues have arisen along with CBIR- privacy of database and query in- formation. A privacy issue arises when an untrusted party want to access the data of private information of another party. The main challenge is content based search should be performed without revealing the query content and the database information to server or any intruder because according to recent trend Google, face book, icker, and other web applications are unknowingly collecting client interests, which are used for recommendation system, creating political polls, and many others. We are using robust hash values for preventing the query from re- vealing the original content and features of the original content along with client can randomly eliminate certain bit for creating then ambiguity for the server be- cause server has to return all the possible combination of the omitted bits that will create ambiguousness for server, what is returned by database server to the client[17]. Since, both the parties, the client and the server, have only exchanged their hash values therefore privacy of client interest and database content are pro- tected. There is a feature oftunablity, where we can adjust the privacy level. Random projection is the algorithm[18] for calculating the robust hash values. Furthermore, due to returning near duplicates will also lead to weaken the pri- vacy, called as voting attack and due to omitting the bits of robust hash will cost more bandwidth and time. In addition, in our proposed work we are using en- cryption on fetch data by server prevent server's proprietary information and the server's data