This is the first comprehensive survey of membership inference attacks (MIAs) on recommender systems (Rec Sys). MIAs infer whether a user interaction record was used to train a target model, and a successful inference can directly breach privacy—for example, identifying that a purchase record trained a Rec Sys tied to a specific user can reveal that user's special quirks.