Recommender systems are used to predict the interest of a customer on a particular item based on that customer's ratings on other items. Many websites like Amazon and Netflix use such systems to recommend items of potential interest to their customers. In order to improve their recommendation service, the data owners sometimes publicly release all or part of their recommendation data i.e., the ratings of their customers on various items without any person specific detail like the customer names. Still, this released data could suffer from re-identification attacks compromising the customer's privacy.
However, such releases in the past, like the one by Netflix, proved to be fruitful. So, in our work, we propose a technique to publish these recommendation datasets without compromising the privacy of the customers. The goal of this thesis is to provide better privacy and utility than the current solutions.