Search Results for « Differential Privacy »

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.

Abstract In the past decade, differential privacy has seen remarkable success as a rigorous and practical formalization of data privacy. This privacy definition and its divergence based relaxations, however, have several acknowledged weaknesses, either in handling composition of private algorithms or in analysing important primitives like privacy amplification by subsampling. Inspired by the hypothesis testing formulation of privacy, this paper proposes a new relaxation of differential privacy, which we term ‘f-differential privacy’ (f-DP). This notion of privacy has a number of appealing properties and, in particular, avoids difficulties associated with divergence based relaxations. First, f-DP faithfully preserves the hypothesis testing interpretation of differential privacy, thereby making the privacy guarantees easily interpretable. In addition, f-DP allows for lossless reasoning about composition in an algebraic fashion. Moreover, we provide a powerful technique to import existing results proven for the original differential privacy definition to f-DP and, as an application of this technique, obtain a simple and easy-to-interpret theorem of privacy amplification by subsampling for f-DP. In addition to the above findings, we introduce a canonical single-parameter family of privacy notions within the f-DP class that is referred to as ‘Gaussian differential privacy’ (GDP), defined based on hypothesis testing of two shifted Gaussian distributions. GDP is the focal privacy definition among the family of f-DP guarantees due to a central limit theorem for differential privacy that we prove. More precisely, the privacy guarantees of any hypothesis testing based definition of privacy (including the original differential privacy definition) converges to GDP in the limit under composition. We also prove a Berry–Esseen style version of the central limit theorem, which gives a computationally inexpensive tool for tractably analysing the exact composition of private algorithms. Taken together, this collection of attractive properties render f-DP a mathematically coherent, analytically tractable and versatile framework for private data analysis. Finally, we demonstrate the use of the tools we develop by giving an improved analysis of the privacy guarantees of noisy stochastic gradient descent.

Boosting is a general method for improving the accuracy of learning algorithms. We use boosting to construct improved privacy-pre serving synopses of an input database. These are data structures that yield, for a given set Q of queries over an input database, reasonably accurate estimates of the responses to every query in Q, even when the number of queries is much larger than the number of rows in the database. Given a base synopsis generator that takes a distribution on Q and produces a "weak" synopsis that yields "good" answers for a majority of the weight in Q, our Boosting for Queries algorithm obtains a synopsis that is good for all of Q. We ensure privacy for the rows of the database, but the boosting is performed on the queries. We also provide the first synopsis generators for arbitrary sets of arbitrary low-sensitivity queries, i.e., queries whose answers do not vary much under the addition or deletion of a single row. In the execution of our algorithm certain tasks, each incurring some privacy loss, are performed many times. To analyze the cumulative privacy loss, we obtain an O(ε2) bound on the expected privacy loss from a single e-differentially private mechanism. Combining this with evolution of confidence arguments from the literature, we get stronger bounds on the expected cumulative privacy loss due to multiple mechanisms, each of which provides e-differential privacy or one of its relaxations, and each of which operates on (potentially) different, adaptively chosen, databases.

Machine learning techniques based on neural networks are achieving remarkable results in a wide variety of domains. Often, the training of models requires large, representative datasets, which may be crowdsourced and contain sensitive information. The models should not expose private information in these datasets. Addressing this goal, we develop new algorithmic techniques for learning and a refined analysis of privacy costs within the framework of differential privacy. Our implementation and experiments demonstrate that we can train deep neural networks with non-convex objectives, under a modest privacy budget, and at a manageable cost in software complexity, training efficiency, and model quality.

One goal of statistical privacy research is to construct a data release mechanism that protects individual privacy while preserving information content. An example is a random mechanism that takes an input database X and outputs a random database Z according to a distribution Qn(⋅|X). Differential privacy is a particular privacy requirement developed by computer scientists in which Qn(⋅|X) is required to be insensitive to changes in one data point in X. This makes it difficult to infer from Z whether a given individual is in the original database X. We consider differential privacy from a statistical perspective. We consider several data-release mechanisms that satisfy the differential privacy requirement. We show that it is useful to compare these schemes by computing the rate of convergence of distributions and densities constructed from the released data. We study a general privacy method, called the exponential mechanism, introduced by McSherry and Talwar (2007). We show that the accuracy of this method is intimately linked to the rate at which the probability that the empirical distribution concentrates in a small ball around the true distribution.

We consider the noise complexity of differentially private mechanisms in the setting where the user asks d linear queries f:Rn -> R non-adaptively. Here, the database is represented by a vector in R and proximity between databases is measured in the l1-metric. We show that the noise complexity is determined by two geometric parameters associated with the set of queries. We use this connection to give tight upper and lower bounds on the noise complexity for any d ≤ n. We show that for d random linear queries of sensitivity 1, it is necessary and sufficient to add l2-error Θ(min d√d/ε,d√(log (n/d))/ε) to achieve ε-differential privacy. Assuming the truth of a deep conjecture from convex geometry, known as the Hyperplane conjecture, we can extend our results to arbitrary linear queries giving nearly matching upper and lower bounds.

Differential privacy is a popular privacy model within the research community because of the strong privacy guarantee it offers, namely that the presence or absence of any individual in a data set does not significantly influence the results of analyses on the data set. However, enforcing this strict guarantee in practice significantly distorts data and/or limits data uses, thus diminishing the analytical utility of the differentially private results. In an attempt to address this shortcoming, several relaxations of differential privacy have been proposed that trade off privacy guarantees for improved data utility. In this paper, we argue that the standard formalization of differential privacy is stricter than required by the intuitive privacy guarantee it seeks. In particular, the standard formalization requires indistinguishability of results between any pair of neighbor data sets, while indistinguishability between the actual data set and its neighbor data sets should be enough. This limits the data controller's ability to adjust the level of protection to the actual data, hence resulting in significant accuracy loss. In this respect, we propose individual differential privacy, an alternative differential privacy notion that offers the same privacy guarantees as standard differential privacy to individuals (even though not to groups of individuals). This new notion allows the data controller to adjust the distortion to the actual data set, which results in less distortion and more analytical accuracy. We propose several mechanisms to attain individual differential privacy and we compare the new notion against standard differential privacy in terms of the accuracy of the analytical results.

Sequential querying of differentially private mechanisms degrades the overall privacy level. In this paper, we answer the fundamental question of characterizing the level of overall privacy degradation as a function of the number of queries and the privacy levels maintained by each privatization mechanism. Our solution is complete: we prove an upper bound on the overall privacy level and construct a sequence of privatization mechanisms that achieves this bound. The key innovation is the introduction of an operational interpretation of differential privacy (involving hypothesis testing) and the use of a data processing inequality along with its converse. Our result improves over the state of the art, and has immediate connections to several problems studied in the literature.

Differential privacy is a precise mathematical constraint meant to ensure privacy of individual pieces of information in a database even while queries are being answered about the aggregate. Intuitively, one must come to terms with what differential privacy does and does not guarantee. For example, the definition prevents a strong adversary who knows all but one entry in the database from further inferring about the last one. This strong adversary assumption can be overlooked, resulting in misinterpretation of the privacy guarantee of differential privacy. Herein we give an equivalent definition of privacy using mutual information that makes plain some of the subtleties of differential privacy. The mutual-information differential privacy is in fact sandwiched between ε-differential privacy and (ε,δ)-differential privacy in terms of its strength. In contrast to previous works using unconditional mutual information, differential privacy is fundamentally related to conditional mutual information, accompanied by a maximization over the database distribution. The conceptual advantage of using mutual information, aside from yielding a simpler and more intuitive definition of differential privacy, is that its properties are well understood. Several properties of differential privacy are easily verified for the mutual information alternative, such as composition theorems.