Data mining with differential privacy

Friedman, Arik; Schuster, Assaf

doi:10.1145/1835804.1835868

Cited by 384 publications

(314 citation statements)

References 19 publications

Supporting

Mentioning

311

Contrasting

Unclassified

Order By: Relevance

“…As we will see in Section 5, our results can also be adapted to cover the case of unbounded-DP, thus further extending their applicability to other use-cases of differential privacy. Examples of settings where DP mechanisms have been proposed, and yet an adversary with incomplete background knowledge appears reasonable, can be found in location privacy [1] or data mining [7] for instance.…”

Section: Discussionmentioning

confidence: 99%

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

Section: Discussionmentioning

confidence: 99%

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

show abstract

“…• Comparing the quality of the patterns discovered in data before and after applying privacy-preserving techniques to the data (Fletcher & Islam 2014, 2015a, Friedman & Schuster 2010; and…”

Section: Problem Statementmentioning

confidence: 99%

Comparing sets of patterns with the Jaccard index

Fletcher

Islam

2018

AJIS

View full text Add to dashboard Cite

The ability to extract knowledge from data has been the driving force of Data Mining since its inception, and of statistical modelling long before even that. Actionable knowledge often takes the form of patterns, where a set of antecedents can be used to infer a consequent. In this paper we offer a solution to the problem of comparing different sets of patterns. Our solution allows comparisons between sets of patterns that were derived from different techniques (such as different classification algorithms), or made from different samples of data (such as temporal data or data perturbed for privacy reasons). We propose using the Jaccard index to measure the similarity between sets of patterns by converting each pattern into a single element within the set. Our measure focuses on providing conceptual simplicity, computational simplicity, interpretability, and wide applicability. The results of this measure are compared to prediction accuracy in the context of a real-world data mining scenario.

show abstract

“…Two major areas have by now clearly emerged: Quantitative Information Flow (qif) [8,19,5,6,9,10,26] and Differential Privacy (dp) [13,14,21,22,16,17]. As discussed in [4], qif is mainly concerned with quantifying the degree of protection offered against an adversary trying to guess the whole secret; dp is rather concerned with protection of individual bits of the secret, possibly in the presence of background information, like knowledge of the remaining bits.…”

Section: Introductionmentioning

confidence: 99%

“…This is an instance of the composition attacks which are well known in the context of dp, where they are thwarted by allotting each user or group of users a privacy budget that limits the overall number of queries to the mechanism; see e.g. [21,16]. For another example, in a de-anonymization scenario similar to [23], [6] shows that gathering information about a target individual can be modeled as collecting multiple observations from a certain randomization mechanism.…”

Section: Introductionmentioning

confidence: 99%

Worst- and Average-Case Privacy Breaches in Randomization Mechanisms

Boreale¹,

Paolini²

2012

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. In a variety of contexts, randomization is regarded as an effective technique to conceal sensitive information. We model randomization mechanisms as information-theoretic channels. Our starting point is a semantic notion of security that expresses absence of any privacy breach above a given level of seriousness , irrespective of any background information, represented as a prior probability on the secret inputs. We first examine this notion according to two dimensions: worst vs. average case, single vs. repeated observations. In each case, we characterize the security level achievable by a mechanism in a simple fashion that only depends on the channel matrix, and specifically on certain measures of "distance" between its rows, like norm-1 distance and Chernoff Information. We next clarify the relation between our worst-case security notion and differential privacy (dp): we show that, while the former is in general stronger, the two coincide if one confines to background information that can be factorised into the product of independent priors over individuals. We finally turn our attention to expected utility, in the sense of Ghosh et al., in the case of repeated independent observations. We characterize the exponential growth rate of any reasonable utility function. In the particular case the mechanism provides -dp, we study the relation of the utility rate with : we offer either exact expressions or upper-bounds for utility rate that apply to practically interesting cases, such as the (truncated) geometric mechanism.

show abstract

Data mining with differential privacy

Cited by 384 publications

References 19 publications

Differential Privacy with Bounded Priors

Differential Privacy with Bounded Priors

Comparing sets of patterns with the Jaccard index

Worst- and Average-Case Privacy Breaches in Randomization Mechanisms

Contact Info

Product

Resources

About