Privacy-Preserving Friendship-Based Recommender Systems

Abstract

Privacy-preserving recommender systems have been an active research topic for many years. However, until today, it is still a challenge to design an efficient solution without involving a fully trusted third party or multiple semitrusted third parties. The key obstacle is the large underlying user populations (i.e. huge input size) in the systems. In this paper, we revisit the concept of friendship-based recommender systems, proposed by Jeckmans et al. and Tang and Wang. These solutions are very promising because recommendations are computed based on inputs from a very small subset of the overall user population (precisely, a user's friends and some randomly chosen strangers). We first clarify the single prediction protocol and Top-n protocol by Tang and Wang, by correcting some flaws and improving the efficiency of the single prediction protocol. We then design a decentralized single protocol by getting rid of the semi-honest service provider. In order to validate the designed protocols, we crawl Twitter and construct two datasets (FMT and 10-FMT) which are equipped with auxiliary friendship information. Based on 10-FMT and MovieLens 100k dataset with simulated friendships, we show that even if our protocols use a very small subset of the datasets, their accuracy can still be equal to or better than some baseline algorithm. Based on these datasets, we further demonstrate that the outputs of our protocols leak very small amount of information of the inputs, and the leakage decreases when the input size increases. We finally show that he single prediction protocol is quite efficient but the Top-n is not. However, we observe that the efficiency of the Top-n protocol can be dramatically improved if we slightly relax the desired security guarantee.

A Survey on Privacy in Social Media

“…Following this strategy, no third parties or external entities need to be trusted by the users to preserve their privacy. Existing approaches use different techniques and mechanisms and could be categorized mainly into three categories: cryptographic-based techniques [6,21,40,74,172], differential privacy-based approaches [66,76,89,120,128,130,166,197,198], and perturbation-based techniques [75,118,146,147,148,151,153,158,183] A group of works focus on providing cryptographic solutions to the problem of secure recommender systems. The approaches do not let the single trusted party have access to everyone's data [6,21,40,74,172].…”

Exposing Private User Behaviors of Collaborative Filtering via Model Inversion Techniques

“…This is arguably a harder to acquire information compared to the item profile, however, we do not consider this as a significant impediment to our attack. Namely, due to the wide spread of social networks and its increasing role in recommendation systems [35,56,60,61], it has become common for users to share their personal (non-sensitive) recommendations through social networks: for example most websites running recommender systems provide the users the option to embed the information to blogs or to share it via Twitter and Facebooks. We can also consider alternative indirect approaches.…”

“…Here, behavior is the purchase or rating histories of users in a recommender system, and can be further divided into public and private behavior. Public behavior is shared through social networks such as blogs and Facebook [35,56,60,61], or through item similarity lists, e.g., Amazon's "Customers who bought this item also bought..." feature [8]. Private behavior is those wished to be kept private such as purchase history of sensitive items; e.g., medical items, sexual movies.…”

“…We further assume the adversary knows nonsensitive items included in the top-l items recommended to the users by the recommender system, i.e., she gets to see the public output of the recommender system. This has been a widely accepted assumption in MI attacks [14,15,21,64], and is also backed up by the spread of sharing (non-sensitive) personal information/behavior via social networks to improve the effectiveness of recommender systems [35,56,60,61], e.g., sharing personal ratings of items on blogs, Facebook, and IMDB.…”

A Research on Collaborative Filtering Based Movie Recommendations: From Neighborhood to Deep Learning Based System

“…Recommender system has been applied to many application areas like Music [4], Friend [5], Trip [6], Movie [7][8] [9], News [10], Social-Media recommendation [11], and many more. Recommending similar movies to the active user as per the likeness is called Movie Recommender System (MRS).…”