Recommender systems have become very popular in recent years and are used in different fields of technology. Memory-based Collaborative Filtering (CF) Recommender System is a quickly progressing study area and proved to be doing well for different types of recommender system. Memory-based CF recommends items based on the entire collection of items or products which have been rated by users previously. CF recommender system tries to provide the most suitable recommendations on products or services based on rating items which can either be explicit or implicit and accuracy of prediction depend on the data provided by users. When CF generates recommendations or predictions on the basis of neighbors in memory known as Memory based CF. In this study, we have discussed a major challenge for the recommender system that has direct access to user data that leads to privacy risk and may become the cause of attacks and other risks. Leading to the above challenges, we have conducted acomprehensive survey of different risks and analyzed them critically. Moreover, we have investigated the existing state-of-the-art approaches which are being used to address such challenges.