Abstract: A privacy protection framework for large-scale content-based information retrieval is proposed, which offers 2 layers of protection. To start with, strong hash values square measure used as queries to avoid uncovering distinctive content or options. Second, the client will favour to exclude bound bits in an exceedingly hash values to additional expand the anomaly for the server. Because of the reduced info, it's computationally tough for the server to grasp the customer's interest. The server has to return the hash values of each single client. The client performs a research at intervals the candidate list to find the simplest match. Since simply hash values square measure changed between client and the server, the privacy of each side is ensured. The thought is to highlight vector into items and list each piece with a sub hash value. Every sub hash price is connected with associate inverted index list. The outcomes demonstrate that the privacy upgrade somewhat enhances the retrieval performance.

Keywords: Multimedia database, indexing, content-based retrieval, data privacy.