2011 IEEE Symposium on Security and Privacy “You Might Also Like:” Privacy Risks of Collaborative Filtering

Citations in the corpus

• Differential Privacy for Growing Databases (2018, background): “Analysis of sensitive data without formal privacy guarantees has led to numerous privacy violations in practice, for tasks such as recommender systems [CKN + 11], targeted advertising [Kor10], data anonymization [NS08], and deep learning [HAPC17].”
• Poisoning Attacks to Graph-Based Recommender Systems (2018, background): “e. 2.2.2 Privacy A‡acks. Two a−acks, i.e., item inference a‹acks and a‹ribute inference a‹acks, were proposed to compromise user privacy in recommender systems. Iteminferenceattacks:Calandrino et al. [4] proposed privacy attacks to infer the items that a target user has rated before, e.g., such item”
• LOGAN: Membership Inference Attacks Against Generative Models (2017, background): “vidual is in a mixture by comparing a distance measure between the individual and the mixture versus the individual and the general population. Other attacks, e.g., on distributed recommender systems [9], infer which inputs cause output changes by looking at temporal patterns of the model. Ateniese ”
• Towards Making Systems Forget with Machine Unlearning (2015, methodology): “For LensKit, we reproduced an existing privacy attack [29].”
• LOGAN: Evaluating Privacy Leakage of Generative Models Using Generative Adversarial Networks (2017, background): “, on distributed recommender systems [7], infer which inputs cause output changes by looking at temporal patterns of the model.”
• Privacy via the Johnson-Lindenstrauss Transform (2012, background): “The reason is that complex algorithms that access private data can be susceptible to either unintended or adversarial attacks [6].”
• Efficient Repair of Polluted Machine Learning Systems via Causal Unlearning (2018, background): “classified as two sub-categories: model inversion [12, 23] where an attacker infers training data samples based on the learning model, and data evasion [7, 13, 29, 44, 45] where an attacker crafts samples to evade the learning model.”
• Secure Multi-party Differential Privacy (2015, background): “Such attacks have revealed the medical record of a former governor of Massachusetts [37], the purchase history of Amazon users[7], genomic information [25], and movie viewing history of Netflix users [33].”
• The Unintended Consequences of Overfitting: Training Data Inference Attacks (2017, background): “Calandrino et al. [44] showed that temporal changes in recommendations given by collaborative filtering methods can reveal the inputs that caused those changes.”
• Protecting privacy when mining and sharing user data (2012, background): “[31] demonstrate that algorithms that recommend products based on friends’ purchases have very practical privacy concerns.”
• Testing the Lipschitz Property over Product Distributions with Applications to Data Privacy (2013, background): “rich in user information, it poses serious privacy concerns about the users’ information contained in the data [23,6].”
• Membership Inference Attacks Against Machine Learning Models (2016, background): “Other attacks on machine learning include [7], where the adversary exploits changes in the outputs of a collaborative recommender system to infer inputs that caused these changes.”