Coupled-Worlds Privacy: Exploiting Adversarial Uncertainty in Statistical Data Privacy

Bassily, Raef; Groce, Adam; Katz, Jonathan; Smith, Adam

doi:10.1109/focs.2013.54

Cited by 60 publications

(93 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Several other works, most notably the Pufferfish and the coupled-worlds frameworks [15], [16], propose different stability constraints on the output distribution of privacypreserving mechanisms. Although they differ in what distributions are compared, their notion of closeness is the same as in (ǫ, δ)-DP.…”

Section: Differential Privacy and Its Flavorsmentioning

confidence: 99%

Rényi Differential Privacy

Mironov

2017

2017 IEEE 30th Computer Security Foundations Symposium (CSF)

868

1,023

View full text Add to dashboard Cite

Abstract-We propose a natural relaxation of differential privacy based on the Rényi divergence. Closely related notions have appeared in several recent papers that analyzed composition of differentially private mechanisms. We argue that the useful analytical tool can be used as a privacy definition, compactly and accurately representing guarantees on the tails of the privacy loss.We demonstrate that the new definition shares many important properties with the standard definition of differential privacy, while additionally allowing tighter analysis of composite heterogeneous mechanisms.

show abstract

Section: Differential Privacy and Its Flavorsmentioning

confidence: 99%

Rényi Differential Privacy

Mironov

2017

2017 IEEE 30th Computer Security Foundations Symposium (CSF)

868

1,023

View full text Add to dashboard Cite

show abstract

“…Our definition can be seen as an instantiation of this general framework. This is in contrast to other kinds of relaxations of differential privacy, which relax the worst-case assumptions on the prior beliefs of an attacker as in Bassily et al (12), or the worstcase collusion assumptions on collections of data analysts as in Kearns et al (13). Several works have also proposed assigning different differential privacy parameters to different individuals (see, e.g., ref.…”

Section: Significancementioning

confidence: 93%

Private algorithms for the protected in social network search

Kearns

Roth

et al. 2016

Proc. Natl. Acad. Sci. U.S.A.

View full text Add to dashboard Cite

Motivated by tensions between data privacy for individual citizens and societal priorities such as counterterrorism and the containment of infectious disease, we introduce a computational model that distinguishes between parties for whom privacy is explicitly protected, and those for whom it is not (the targeted subpopulation). The goal is the development of algorithms that can effectively identify and take action upon members of the targeted subpopulation in a way that minimally compromises the privacy of the protected, while simultaneously limiting the expense of distinguishing members of the two groups via costly mechanisms such as surveillance, background checks, or medical testing. Within this framework, we provide provably privacy-preserving algorithms for targeted search in social networks. These algorithms are natural variants of common graph search methods, and ensure privacy for the protected by the careful injection of noise in the prioritization of potential targets. We validate the utility of our algorithms with extensive computational experiments on two large-scale social network datasets.data privacy | social networks | counterterrorism T he tension between the useful or essential gathering and analysis of data about citizens and the privacy rights of those citizens is at an historical peak. Perhaps the most striking and controversial recent example is the revelation that US intelligence agencies systemically engage in "bulk collection" of civilian "metadata" detailing telephonic and other types of communication and activities, with the alleged purpose of monitoring and thwarting terrorist activity (1). Other compelling examples abound, including in medicine (patient privacy vs. preventing epidemics), marketing (consumer privacy vs. targeted advertising), and many other domains.Debates about (and models for) data privacy often have an "all or nothing" flavor: privacy guarantees are either provided to every member of a population, or else privacy is deemed to be a failure. This dichotomy is only appropriate if all members of the population have an equal right to, or demand for, privacy. Few would argue that actual terrorists should have such rights, which leads to difficult questions about the balance between protecting the rights of ordinary citizens, and using all available means to prevent terrorism. [A recent National Academies study (2) reached the conclusion that there are not (yet) technological alternatives to bulk collection and analysis of civilian metadata, in the sense that such data are essential in current counterterrorism practices.] A major question is whether and when individual privacy should be sacrificed in service of such societal priorities. Similarly, in the medical domain, epidemics (such as the recent international outbreak of Ebola; ref.3) have raised serious debate about the clear public interest in controlling contagion versus the privacy rights of the infected and those that care for them.The model and results in this paper represent a step toward explicit acknowledgmen...

show abstract

“…Previous works mainly focus on the unbounded-DP case, and thus are not directly applicable to situations where the size of the dataset is public. Furthermore, previously considered adversarial priors are either uniform [13,2] or only allow for a fixed number of known entities [8,15]. Finally, very few results are known on how to design general mechanisms satisfying distributional variants of DP.…”

Section: Relation To Prior Workmentioning

confidence: 99%

“…Works on Differential-Privacy under Sampling [13], Crowd-Blending Privacy [8], Coupled-Worlds Privacy [2] or Outlier Privacy [15] have shown that if sufficiently many users are indistinguishable by a mechanism, and this mechanism operates on a dataset obtained through a robust sampling procedure, differential privacy can be satisfied with only little data perturbation. Our work differs in that we make no assumptions on the indistinguishability of different entities, and that our aim is to guarantee membership privacy rather than differential privacy.…”

Section: Relation To Prior Workmentioning

confidence: 99%

“…The relevance of such a strong adversarial setting has been put into question because it seems unlikely, in a practical data setting, for an adversary to have such a high certainty about all entities. Alternative privacy definitions such as differential-privacy under sampling [13], crowdblending privacy [8], coupled-worlds privacy [2], outlier privacy [15], -privacy [16], or differential identifiability [12] relax the adversarial setting of DP, with the goal of achieving higher utility.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Differential Privacy with Bounded Priors

Tramèr

Huang

Hubaux

et al. 2015

Proceedings of the 22nd ACM SIGSAC Conference on Computer and Communications Security

View full text Add to dashboard Cite

Differential privacy (DP) has become widely accepted as a rigorous definition of data privacy, with stronger privacy guarantees than traditional statistical methods. However, recent studies have shown that for reasonable privacy budgets, differential privacy significantly affects the expected utility. Many alternative privacy notions which aim at relaxing DP have since been proposed, with the hope of providing a better tradeoff between privacy and utility.At CCS'13, Li et al. introduced the membership privacy framework, wherein they aim at protecting against set membership disclosure by adversaries whose prior knowledge is captured by a family of probability distributions. In the context of this framework, we investigate a relaxation of DP, by considering prior distributions that capture more reasonable amounts of background knowledge. We show that for different privacy budgets, DP can be used to achieve membership privacy for various adversarial settings, thus leading to an interesting tradeoff between privacy guarantees and utility.We re-evaluate methods for releasing differentially private χ 2 -statistics in genome-wide association studies and show that we can achieve a higher utility than in previous works, while still guaranteeing membership privacy in a relevant adversarial setting.

show abstract

Coupled-Worlds Privacy: Exploiting Adversarial Uncertainty in Statistical Data Privacy

Cited by 60 publications

References 10 publications

Rényi Differential Privacy

Rényi Differential Privacy

Private algorithms for the protected in social network search

Differential Privacy with Bounded Priors

Contact Info

Product

Resources

About