Robust De-anonymization of Large Sparse Datasets

By Arvind Narayanan, Vitaly Shmatikov. 2008

2,668 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Sociology and Political Science, Privacy & Security

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Sociology and Political Science

In Privacy & Security

Citations in the corpus

Model Inversion Attacks that Exploit Confidence Information and Basic Countermeasures (2015, background): “Narayananan [32] demonstrated that an adversary with some prior knowledge can identify a subscriber’s record in the anonymized Netflix prize dataset.”
Privacy in Deep Learning: A Survey (2020, methodology): “They showed that an adversary with auxiliary knowledge (from the publicly available Internet Movie Database records) about individual subscribers can easily identify the user and uncover potentially sensitive information [49].”
An Economic Analysis of Privacy Protection and Statistical Accuracy as Social Choices (2018, background): “Narayanan and Shmatikov (2008) successfully re-identified a large share of the users in the Netflix Prize data by probabilistically matching to data scraped from IMDb.com, a crowd-sourced movie review site.”
Differentially Private Federated Learning: A Client Level Perspective (2017, background): “Unfortunately, anonymization alone often is not sufficient [7, 2] and standard machine learning approaches largely disregard privacy aspects.”
Prochlo: Strong Privacy for Analytics in the Crowd (2017, background): “Famously, Narayanan and Shmatikov [54] managed to deanonymize users in the challenge’s large corpus of training data by exploiting auxiliary data and the linkability of users’ movie preferences.”
BLENDER: Enabling Local Search with a Hybrid Differential Privacy Model (2017, background): “In the last decade, we have witnessed scores of adhoc approaches that have turned out to be inadequate for protecting privacy [33, 23].”
Scalable Private Learning with PATE (2018, background): “This is not the case of all privacy definitions, as demonstrated by successful attacks against anonymization techniques (Aggarwal, 2005; Narayanan & Shmatikov, 2008; Bindschaedler et al., 2017).”
Differentially Private Password Frequency Lists (2016, background): “a) Privacy Models: De-anonymization attacks [28], [35] highlight the problems with heuristic approaches to data sanitization (e.”
Security Evaluation of Support Vector Machines in Adversarial Environments (2014, background): “Given an adversary with such strong knowledge and capabilities as described above, it may seem difficult to provide effective countermeasures particularly considering the complication of abundant access to side information that is often used in publicized privacy attacks [51, 64].”
Social Data: Biases, Methodological Pitfalls, and Ethical Boundaries (2019, neutral): “Prominent releases of anonymized data, such as for the Netﬂix prize and the AOL’s Search History Database, were later found to provide inadequate protection to users (Barbaro et al., 2006; Narayanan and Shmatikov, 2008).”
Differentially Private Empirical Risk Minimization (2009, neutral): “Such attacks have been demonstrated by various works, and are possible in many realistic settings, such as when an adversary has side information (Sweeney, 1997; Narayanan and Shmatikov, 2008; Ganta et al., 2008), and when the data has structural properties (Backstrom et al., 2007), among others.”
Large image datasets: A pyrrhic win for computer vision? (2020, background): “Institutions, academia, and industry alike, amass millions of images of people without consent and often for unstated purposes under the guise of anonymization, a claim that is both ephemeral [57, 68] and vacuous [30].”
Private traits and attributes are predictable from digital records of human behavior (2013, background): “However, the widespread availability of extensive records of individual behavior, together withthe desiretolearnmoreaboutcustomersandcitizens,presents serious challenges related to privacy and data ownership (4, 5).”
Universally utility-maximizing privacy mechanisms (2008, background): “Unfortunately, publication schemes initially thought to be “private” have succumbed to privacy attacks [19, 17, 1], highlighting the urgent need for mechanisms that are provably private.”
Privacy via the Johnson-Lindenstrauss Transform (2012, background): “Approaches resorting to user id anonymization while keeping the data unchanged have been shown to be badly insuﬃcient to preserve privacy [2, 29].”