Currently, there are two approaches to limit disclosure of databases containing private information: 1) altering the data before disclosure and 2) learning results without revealing the data. In this talk, I present a watermarking scheme that can be considered as a third approach or a complement to the two existing approaches. The scheme uses watermarking to identify the owner of private information and consists of a watermark, encoder, decoder and a detection algorithm. An analysis of the scheme is also provided.