The cost of privacy: destruction of data-mining utility in anonymized data publishing

By Justin Brickell, Vitaly Shmatikov. 2008

404 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Sociology and Political Science

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

Citations in the corpus

Privacy in Pharmacogenetics: An End-to-End Case Study of Personalized Warfarin Dosing (2014, neutral): “Brickell and Shmatikov [6] found strong evidence for a tradeoff in attribute privacy and predictive performance in common data mining tasks when k-anonymity, `-diversity, and t-closeness are applied before releasing a full dataset.”
Data Synthesis based on Generative Adversarial Networks (2018, background): “δ-disclosure is also a concept to protect sensitive attributes from re-identification attacks [14].”
Anonymization of location data does not work: a large-scale measurement study (2011, background): “Random sampling or random noise insertion impacts the utility of data and the tradeoff between utility and privacy need to be carefully considered [21, 17, 2].”
Minimax Filter: Learning to Preserve Privacy from Inference Attacks (2016, methodology): “Regarding the use of risks in this paper, classiﬁcation error-based quantities have been suggested in the literature (Iyengar, 2002; Brickell and Shmatikov, 2008; Li and Li, 2009).”
Stolen Memories: Leveraging Model Memorization for Calibrated White-Box Membership Inference (2019, background): “Many compelling applications of machine learning involve the collection and processing of sensitive personal data, giving rise to concerns about privacy [2, 4, 7, 10, 11, 26, 33, 38, 45, 46].”
On the tradeoff between privacy and utility in data publishing (2009, methodology): “In fact, the direct-comparison methodology [5] underestimates the seriousness of privacy loss, as it uses average privacy loss among all individuals.”
Privacy in Search Logs (2009, methodology): “[4] measure the utility with respect to data mining tasks and Kifer and Gehrke [11] develop specific techniques to boost utility with respect to log-linear models.”
Just What do You Think You're Doing, Dave?' A Checklist for Responsible Data Use in NLP (2021, neutral): “There is also a fundamental tradeoff between data utility and the privacy it affords: any anonymisation technique necessarily removes information from data, which may signiﬁcantly decrease its utility (Brickell and Shmatikov, 2008; Zang and Bolot, 2011).”
Utility-Privacy Tradeoffs in Databases: An Information-Theoretic Approach (2011, background): “original database and those of the perturbed database. The general concept of utility as a measure of the approximation to an underlying (but undisclosed) quantity is a fertile area of research [30], [31], [34], [35]. However, these measures have not been customized for the context of privacy enhanc”
Slicing: A New Approach to Privacy Preserving Data Publishing (2009, background): “Every column contains a subset [4] of attributes.”
Differential Privacy for Deep and Federated Learning: A Survey (2022, background): “Nevertheless, k − anonymization performs poorly on the anonymization of a high-dimensional dataset and does not provide strong protection against attribute disclosure [14], [15].”
Publishing Search Logs—A Comparative Study of Privacy Guarantees (2009, methodology): “To avoid this drawback, Brickell and Shmatikov [6] measure the utility with respect to data mining tasks and they take the actual classification error of an induced classifier as their utility metric.”
Privacy Risk in Machine Learning: Analyzing the Connection to Overfitting (2017, background): “The fact that many compelling applications of this technology involve the collection and processing of sensitive personal data has given rise to concerns about privacy [1], [2], [3], [4], [5], [6], [7], [8], [9].”
Differentially Private Mixture of Generative Neural Networks (2017, background): “k-anonymity [41] aims to protect data by generalizing and suppressing certain identifying attributes, however, it does not work well on high-dimensional datasets [3], [9].”
The Unintended Consequences of Overfitting: Training Data Inference Attacks (2017, background): “The fact that many compelling applications of this technology involve the collection and processing of sensitive personal data has given rise to concerns about privacy [1, 2, 3, 4, 5, 6, 7, 8, 9].”