With the explosive growth of digital data, recommendation systems (RSs) play a crucial role in alleviating the problem of information overload. Implicit feedback data has become the primary data source for training recommendation models because of its richness and ease of collection. Leveraging such data for personalized recommendation services requires a large amount of user historical interactions, which poses serious privacy risks. Differential privacy (DP) has been integrated into implicit recommendation algorithms to protect user privacy. However, due to the inherent characteristics of implicit feedback, the current studies still have certain deficiencies in terms of data utility and privacy level. To this end, this article proposes a novel differentially private implicit recommendation algorithm. It integrates the Bayesian personalized ranking (BPR) matrix factorization (MF) with the Gaussian mechanism in Rényi DP (RDP) and designs an optimization strategy based on the binary index tree (BIT) to alleviate the cumulative errors. The proposed method not only can effectively capture the user preferences from sparse implicit feedback data by maximizing the posterior probability of rankings but also can more precisely manage the privacy budget allocation according to the query matrix. The theoretical analyses prove that the proposed method can satisfy the privacy guarantee and give the upper bound of privacy loss. The experimental results show that our method outperforms several existing advanced methods. It achieves a maximum performance improvement of 6.5% and 6.9% on Hit Rate (HR@10) and Normalized Discounted Cumulative Gain (NDCG@10) at a low privacy budget, which indicates that it can provide good recommendation quality while ensuring a strict privacy level.

In recommendation systems, the interactions between users and items are influenced by two factors: the user's conformity towards popular items and the user's real interest. Training individual user embeddings and item embeddings to capture these two factors can effectively improve the accuracy of recommendations. However, recommendation systems often exchange item embeddings with third-party servers, which may expose sensitive information to malicious attackers. Specifically, attackers can infer sensitive user information based on published item embeddings and partial public user information. In this paper, we first design a model inversion attack to analyze the influence of conformity item embeddings and interest item embeddings on privacy. This analysis reveals that different item embeddings have varying resistances against inversion attack. Based on the resistance levels of the two item embeddings, we propose a novel adaptive differential privacy protection method that enhances resistance against model inversion attacks while ensuring recommendation accuracy. We conduct experiments on three real datasets, and the results demonstrate the outstanding performance of our method in terms of both recommendation accuracy and resistance to inversion attack.

Most previous studies on matrix factorization (MF)-based collaborative filtering (CF) have focused solely on user rating information for predicting recommendations. However, to further enhance the performance of recommender systems (RSs), it is important to also consider review information and rating reliability in the model. This article proposes a new probabilistic MF (PMF)-based CF method that integrates multiple information sources to provide reliable predictions. First, we introduce a sentiment-based PMF to handle user reviews and fit the normalized sentiment information obtained from our previously proposed sentiment analysis method. We also consider the helpfulness of reviews to highlight their reliability and effectiveness in this model. Subsequently, our proposed noise detection method is adopted to determine the reliability of user ratings, and then the rating matrix is transformed into a binary reliability matrix. A rating reliability-based PMF through Bernoulli distribution is then proposed to factorize it. To effectively integrate three types of information (ratings, reviews, and rating reliability) into a PMF procedure, we design a weight matrix using the proposed weighting strategy to generate a set of comprehensive prediction ratings with corresponding reliability probabilities. Experiments on four Amazon datasets demonstrate that our model outperforms comparison methods in terms of comprehensive evaluation.