How to hide the elephant- or the donkey- in the room: Practical privacy against statistical inference for large data

Salamatian, Salman; Zhang, Amy; Calmon, Flávio P.; Bhamidipati, Sandilya; Fawaz, Nadia; Kveton, Branislav; Oliveira, Pedro Felipe Alves de; Taft, Nina

doi:10.1109/globalsip.2013.6736867

Cited by 59 publications

(50 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For an overview, we refer the reader to [1], [9] and the references therein. The privacy against statistical inference framework considered here was further studied in [13], [14]. The results presented in this paper are closely connected to the study of hypercontractivity coefficients and strong data processing results, such as in [6], [7], [15]- [17].…”

Section: A Related Workmentioning

confidence: 73%

Fundamental limits of perfect privacy

Calmon

Makhdoumi

Médard

2015

2015 IEEE International Symposium on Information Theory (ISIT)

View full text Add to dashboard Cite

Abstract-"To be considered for an 2015 IEEE Jack Keil Wolf ISIT Student Paper Award." We investigate the problem of intentionally disclosing information about a set of measurement points X (useful information), while guaranteeing that little or no information is revealed about a private variable S (private information). Given that S and X are drawn from a finite set with joint distribution pS,X , we prove that a non-trivial amount of useful information can be disclosed while not disclosing any private information if and only if the smallest principal inertia component of the joint distribution of S and X is 0. This fundamental result characterizes when useful information can be privately disclosed for any privacy metric based on statistical dependence. We derive sharp bounds for the tradeoff between disclosure of useful and private information, and provide explicit constructions of privacy-assuring mappings that achieve these bounds.

show abstract

Section: A Related Workmentioning

confidence: 73%

Fundamental limits of perfect privacy

Calmon

Makhdoumi

Médard

2015

2015 IEEE International Symposium on Information Theory (ISIT)

View full text Add to dashboard Cite

show abstract

“…Closer to our setting, Kosinski et al [25] show that several personality traits, including political views, sexual orientation, and drug use can be accurately predicted from Facebook "likes", while Weinsberg et al [44] show that gender can be inferred from movie ratings with close to 80% accuracy. Salamatian et al [38] also show that political views can be inferred with confidence above 71% by using only a user's ratings to her 5 most-watched TV shows. Privacy-Preserving Data Mining and Information-Theoretic Models.…”

Section: Related Workmentioning

confidence: 99%

“…We evaluate our methods on three datasets: Politics-and-TV, Movielens and Flixster. Politics-and-TV (PTV) [38] is a ratings dataset that includes 5-star ratings of users to 50 TV-shows and, in addition, each user's political affiliation (Democrat or Republican) and gender. To make it dense, we consider only users that rate over 40 items, resulting in 365 users; 280 provide ratings to all 50 TV shows.…”

Section: Methodsmentioning

confidence: 99%

“…To prevent inference of Y from the release, one can apply a distortion f (X) on X; the goal is then to find the minimal distortion so that the mutual information between f (X) and Y is below a threshold. This problem was originally addressed in the asymptotic regime [39,45], while a series of recent works study it in a non-asymptotic setting [7,28,36,38]. Broadly speaking, our work can be cast in this framework by treating a user's ratings as X, her private feature as Y , and employing a correlation structure between them as specified by matrix factorization (namely, (7)).…”

Section: Related Workmentioning

confidence: 99%

“…Recent studies report many examples where strong correlations have been found: attributes successfully inferred from ratings or "likes" include political affiliation [25,38], sexual orientation [25], age [44], gender [38,44], and even drug use [25]. Yet more privacy threats have been extensively documented in literature (see, e.g., [3,30,31,34,35]).…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Privacy tradeoffs in predictive analytics

IoannidisStratis

MontanariAndrea

WeinsbergUdi

et al. 2014

SIGMETRICS Perform. Eval. Rev.

Self Cite

View full text Add to dashboard Cite

Online services routinely mine user data to predict user preferences, make recommendations, and place targeted ads. Recent research has demonstrated that several private user attributes (such as political affiliation, sexual orientation, and gender) can be inferred from such data. Can a privacy-conscious user benefit from personalization while simultaneously protecting her private attributes? We study this question in the context of a rating prediction service based on matrix factorization. We construct a protocol of interactions between the service and users that has remarkable optimality properties: it is privacy-preserving, in that no inference algorithm can succeed in inferring a user's private attribute with a probability better than random guessing; it has maximal accuracy, in that no other privacy-preserving protocol improves rating prediction; and, finally, it involves a minimal disclosure, as the prediction accuracy strictly decreases when the service reveals less information. We extensively evaluate our protocol using several rating datasets, demonstrating that it successfully blocks the inference of gender, age and political affiliation, while incurring less than 5% decrease in the accuracy of rating prediction.

show abstract

Tracking and Personalization

Masood

Berkovsky

Kâafar

2021

Modern Socio-Technical Perspectives on Privacy

View full text Add to dashboard Cite

This chapter studies the relationship between two important, often conflicting paradigms of online services: personalization and tracking. The chapter initially focuses on the categories and levels of online personalization, briefly overviewing algorithmic methods applied to achieve these. Then, the chapter turns to online tracking specific to mobile and web technologies, as well as the more advanced behavioral tracking. Following this, the chapter ties the streams of personalization and tracking together and discusses various aspects of their relationships, including the currently deployed tracking methods for personalization. Privacy implications of personalization via online tracking, highlighted by organizations and researchers, are also illustrated. Lastly, this chapter discusses the ways to balance personalization benefits and privacy concerns. This includes the state-of-the-art practices, current challenges, and practical recommendations for system developers willing to strike this balance.

show abstract

How to hide the elephant- or the donkey- in the room: Practical privacy against statistical inference for large data

Cited by 59 publications

References 10 publications

Fundamental limits of perfect privacy

Fundamental limits of perfect privacy

Privacy tradeoffs in predictive analytics

Tracking and Personalization

Contact Info

Product

Resources

About