arXiv subjects

Catuscia Palamidessi

Publications and source records attributed to Catuscia Palamidessi.

At least 55 records · Page 3Linked to original sources

arXiv2022

Reconstruction of the distribution of sensitive data under free-will privacy

The local privacy mechanisms, such as k-RR, RAPPOR, and the geo-indistinguishability ones, have become quite popular thanks to the fact that the obfuscation can be effectuated at the users end, thus avoiding the need of a trusted third party. Another important advantage is that each data point is sanitized independently from the others, and therefore different users may use different levels of obfuscation depending on their privacy requirements, or they may even use entirely different mechanisms depending on the services they are trading their data for. A challenging requirement in this setting is to construct the original distribution on the users sensitive data from their noisy versions. Existing techniques can only estimate that distribution separately on each obfuscation schema and corresponding noisy data subset. But the smaller are the subsets, the more imprecise the estimations are. In this paper we study how to avoid the subsets-fractioning problem when combining local privacy mechanisms, thus recovering an optimal utility. We focus on the estimation of the original distribution, and on the two main methods to estimate it: the matrix-inversion method and the iterative Bayes update. We consider various cases of combination of local privacy mechanisms, and compare the flexibility and the performance of the two methods.

cs.CR↗

arXiv2022

An Incentive Mechanism for Trading Personal Data in Data Markets

With the proliferation of the digital data economy, digital data is considered as the crude oil in the twenty-first century, and its value is increasing. Keeping pace with this trend, the model of data market trading between data providers and data consumers, is starting to emerge as a process to obtain high-quality personal information in exchange for some compensation. However, the risk of privacy violations caused by personal data analysis hinders data providers' participation in the data market. Differential privacy, a de-facto standard for privacy protection, can solve this problem, but, on the other hand, it deteriorates the data utility. In this paper, we introduce a pricing mechanism that takes into account the trade-off between privacy and accuracy. We propose a method to induce the data provider to accurately report her privacy price and, we optimize it in order to maximize the data consumer's profit within budget constraints. We show formally that the proposed mechanism achieves these properties, and also, validate them experimentally.

cs.GT↗

arXiv2022

Establishing the Price of Privacy in Federated Data Trading

Personal data is becoming one of the most essential resources in today's information-based society. Accordingly, there is a growing interest in data markets, which operate data trading services between data providers and data consumers. One issue the data markets have to address is that of the potential threats to privacy. Usually some kind of protection must be provided, which generally comes to the detriment of utility. A correct pricing mechanism for private data should therefore depend on the level of privacy. In this paper, we propose a model of data federation in which data providers, who are, generally, less influential on the market than data consumers, form a coalition for trading their data, simultaneously shielding against privacy threats by means of differential privacy. Additionally, we propose a technique to price private data, and an revenue-distribution mechanism to distribute the revenue fairly in such federation data trading environments. Our model also motivates the data providers to cooperate with their respective federations, facilitating a fair and swift private data trading process. We validate our result through various experiments, showing that the proposed methods provide benefits to both data providers and consumers.

cs.GT↗

arXiv2022

Causal Discovery for Fairness

It is crucial to consider the social and ethical consequences of AI and ML based decisions for the safe and acceptable use of these emerging technologies. Fairness, in particular, guarantees that the ML decisions do not result in discrimination against individuals or minorities. Identifying and measuring reliably fairness/discrimination is better achieved using causality which considers the causal relation, beyond mere association, between the sensitive attribute (e.g. gender, race, religion, etc.) and the decision (e.g. job hiring, loan granting, etc.). The big impediment to the use of causality to address fairness, however, is the unavailability of the causal model (typically represented as a causal graph). Existing causal approaches to fairness in the literature do not address this problem and assume that the causal model is available. In this paper, we do not make such assumption and we review the major algorithms to discover causal relations from observable data. This study focuses on causal discovery and its impact on fairness. In particular, we show how different causal discovery approaches may result in different causal models and, most importantly, how even slight differences between causal models can have significant impact on fairness/discrimination conclusions. These results are consolidated by empirical analysis using synthetic and standard fairness benchmark datasets. The main goal of this study is to highlight the importance of the causal discovery step to appropriately address fairness using causality.

cs.AI↗

arXiv2022

Survey on Causal-based Machine Learning Fairness Notions

Addressing the problem of fairness is crucial to safely use machine learning algorithms to support decisions with a critical impact on people's lives such as job hiring, child maltreatment, disease diagnosis, loan granting, etc. Several notions of fairness have been defined and examined in the past decade, such as statistical parity and equalized odds. The most recent fairness notions, however, are causal-based and reflect the now widely accepted idea that using causality is necessary to appropriately address the problem of fairness. This paper examines an exhaustive list of causal-based fairness notions and study their applicability in real-world scenarios. As the majority of causal-based fairness notions are defined in terms of non-observable quantities (e.g., interventions and counterfactuals), their deployment in practice requires to compute or estimate those quantities using observational data. This paper offers a comprehensive report of the different approaches to infer causal quantities from observational data including identifiability (Pearl's SCM framework) and estimation (potential outcome framework). The main contributions of this survey paper are (1) a guideline to help selecting a suitable fairness notion given a specific real-world scenario, and (2) a ranking of the fairness notions according to Pearl's causation ladder indicating how difficult it is to deploy each notion in practice.

cs.LG↗

arXiv2022

Identifiability of Causal-based Fairness Notions: A State of the Art

Machine learning algorithms can produce biased outcome/prediction, typically, against minorities and under-represented sub-populations. Therefore, fairness is emerging as an important requirement for the large scale application of machine learning based technologies. The most commonly used fairness notions (e.g. statistical parity, equalized odds, predictive parity, etc.) are observational and rely on mere correlation between variables. These notions fail to identify bias in case of statistical anomalies such as Simpson's or Berkson's paradoxes. Causality-based fairness notions (e.g. counterfactual fairness, no-proxy discrimination, etc.) are immune to such anomalies and hence more reliable to assess fairness. The problem of causality-based fairness notions, however, is that they are defined in terms of quantities (e.g. causal, counterfactual, and path-specific effects) that are not always measurable. This is known as the identifiability problem and is the topic of a large body of work in the causal inference literature. This paper is a compilation of the major identifiability results which are of particular relevance for machine learning fairness. The results are illustrated using a large number of examples and causal graphs. The paper would be of particular interest to fairness researchers, practitioners, and policy makers who are considering the use of causality-based fairness notions as it summarizes and illustrates the major identifiability results

cs.LG↗

arXiv2022

Machine learning fairness notions: Bridging the gap with real-world applications

Fairness emerged as an important requirement to guarantee that Machine Learning (ML) predictive systems do not discriminate against specific individuals or entire sub-populations, in particular, minorities. Given the inherent subjectivity of viewing the concept of fairness, several notions of fairness have been introduced in the literature. This paper is a survey that illustrates the subtleties between fairness notions through a large number of examples and scenarios. In addition, unlike other surveys in the literature, it addresses the question of: which notion of fairness is most suited to a given real-world scenario and why? Our attempt to answer this question consists in (1) identifying the set of fairness-related characteristics of the real-world scenario at hand, (2) analyzing the behavior of each fairness notion, and then (3) fitting these two elements to recommend the most suitable fairness notion in every specific setup. The results are summarized in a decision diagram that can be used by practitioners and policymakers to navigate the relatively large catalog of ML.

cs.LG↗

arXiv2022

Tight Differential Privacy Blanket for Shuffle Model

With the recent bloom of focus on digital economy, the importance of personal data has seen a massive surge of late. Keeping pace with this trend, the model of data market is starting to emerge as a process to obtain high-quality personal information in exchange of incentives. To have a formal guarantee to protect the privacy of the sensitive data involved in digital economy, \emph{differential privacy (DP)} is the go-to technique, which has gained a lot of attention by the community recently. However, it is essential to optimize the privacy-utility trade-off by ensuring the highest level of privacy protection is ensured while preserving the utility of the data. In this paper, we theoretically derive sufficient and necessary conditions to have tight $(ε,\,δ)$-DP blankets for the shuffle model, which, to the best of our knowledge, have not been proven before, and, thus, characterize the best possible DP protection for shuffle models which can be implemented in data markets to ensure privacy-preserving trading of digital economy.

cs.CR↗

arXiv2022

Universal Optimality and Robust Utility Bounds for Metric Differential Privacy

We study the privacy-utility trade-off in the context of metric differential privacy. Ghosh et al. introduced the idea of universal optimality to characterise the best mechanism for a certain query that simultaneously satisfies (a fixed) $ε$-differential privacy constraint whilst at the same time providing better utility compared to any other $ε$-differentially private mechanism for the same query. They showed that the Geometric mechanism is "universally optimal" for the class of counting queries. On the other hand, Brenner and Nissim showed that outside the space of counting queries, and for the Bayes risk loss function, no such universally optimal mechanisms exist. In this paper we use metric differential privacy and quantitative information flow as the fundamental principle for studying universal optimality. Metric differential privacy is a generalisation of both standard (i.e., central) differential privacy and local differential privacy, and it is increasingly being used in various application domains, for instance in location privacy and in privacy preserving machine learning. Using this framework we are able to clarify Nissim and Brenner's negative results, showing (a) that in fact all privacy types contain optimal mechanisms relative to certain kinds of non-trivial loss functions, and (b) extending and generalising their negative results beyond Bayes risk specifically to a wide class of non-trivial loss functions. We also propose weaker universal benchmarks of utility called "privacy type capacities". We show that such capacities always exist and can be computed using a convex optimisation algorithm.

cs.CR↗

arXiv2022

Leveraging Adversarial Examples to Quantify Membership Information Leakage

The use of personal data for training machine learning systems comes with a privacy threat and measuring the level of privacy of a model is one of the major challenges in machine learning today. Identifying training data based on a trained model is a standard way of measuring the privacy risks induced by the model. We develop a novel approach to address the problem of membership inference in pattern recognition models, relying on information provided by adversarial examples. The strategy we propose consists of measuring the magnitude of a perturbation necessary to build an adversarial example. Indeed, we argue that this quantity reflects the likelihood of belonging to the training data. Extensive numerical experiments on multivariate data and an array of state-of-the-art target models show that our method performs comparable or even outperforms state-of-the-art strategies, but without requiring any additional training samples.

cs.LG↗

arXiv2022

Information Leakage Games: Exploring Information as a Utility Function

A common goal in the areas of secure information flow and privacy is to build effective defenses against unwanted leakage of information. To this end, one must be able to reason about potential attacks and their interplay with possible defenses. In this paper, we propose a game-theoretic framework to formalize strategies of attacker and defender in the context of information leakage, and provide a basis for developing optimal defense methods. A novelty of our games is that their utility is given by information leakage, which in some cases may behave in a non-linear way. This causes a significant deviation from classic game theory, in which utility functions are linear with respect to players' strategies. Hence, a key contribution of this paper is the establishment of the foundations of information leakage games. We consider two kinds of games, depending on the notion of leakage considered. The first kind, the QIF-games, is tailored for the theory of quantitative information flow (QIF). The second one, the DP-games, corresponds to differential privacy (DP).

cs.CR↗

arXiv2021

Estimating g-Leakage via Machine Learning

This paper considers the problem of estimating the information leakage of a system in the black-box scenario. It is assumed that the system's internals are unknown to the learner, or anyway too complicated to analyze, and the only available information are pairs of input-output data samples, possibly obtained by submitting queries to the system or provided by a third party. Previous research has mainly focused on counting the frequencies to estimate the input-output conditional probabilities (referred to as frequentist approach), however this method is not accurate when the domain of possible outputs is large. To overcome this difficulty, the estimation of the Bayes error of the ideal classifier was recently investigated using Machine Learning (ML) models and it has been shown to be more accurate thanks to the ability of those models to learn the input-output correspondence. However, the Bayes vulnerability is only suitable to describe one-try attacks. A more general and flexible measure of leakage is the g-vulnerability, which encompasses several different types of adversaries, with different goals and capabilities. In this paper, we propose a novel approach to perform black-box estimation of the g-vulnerability using ML. A feature of our approach is that it does not require to estimate the conditional probabilities, and that it is suitable for a large class of ML algorithms. First, we formally show the learnability for all data distributions. Then, we evaluate the performance via various experiments using k-Nearest Neighbors and Neural Networks. Our results outperform the frequentist approach when the observables domain is large.

cs.CR↗

arXiv2021

On the impossibility of non-trivial accuracy under fairness constraints

One of the main concerns about fairness in machine learning (ML) is that, in order to achieve it, one may have to trade off some accuracy. To overcome this issue, Hardt et al. proposed the notion of equality of opportunity (EO), which is compatible with maximal accuracy when the target label is deterministic with respect to the input features. In the probabilistic case, however, the issue is more complicated: It has been shown that under differential privacy constraints, there are data sources for which EO can only be achieved at the total detriment of accuracy, in the sense that a classifier that satisfies EO cannot be more accurate than a trivial (i.e., constant) classifier. In our paper we strengthen this result by removing the privacy constraint. Namely, we show that for certain data sources, the most accurate classifier that satisfies EO is a trivial classifier. Furthermore, we study the trade-off between accuracy and EO loss (opportunity difference), and provide a sufficient condition on the data source under which EO and non-trivial accuracy are compatible.

cs.LG↗

arXiv2021

DOCTOR: A Simple Method for Detecting Misclassification Errors

Deep neural networks (DNNs) have shown to perform very well on large scale object recognition problems and lead to widespread use for real-world applications, including situations where DNN are implemented as "black boxes". A promising approach to secure their use is to accept decisions that are likely to be correct while discarding the others. In this work, we propose DOCTOR, a simple method that aims to identify whether the prediction of a DNN classifier should (or should not) be trusted so that, consequently, it would be possible to accept it or to reject it. Two scenarios are investigated: Totally Black Box (TBB) where only the soft-predictions are available and Partially Black Box (PBB) where gradient-propagation to perform input pre-processing is allowed. Empirically, we show that DOCTOR outperforms all state-of-the-art methods on various well-known images and sentiment analysis datasets. In particular, we observe a reduction of up to $4\%$ of the false rejection rate (FRR) in the PBB scenario. DOCTOR can be applied to any pre-trained model, it does not require prior information about the underlying dataset and is as simple as the simplest available methods in the literature.

cs.CV↗

arXiv2020

F-BLEAU: Fast Black-box Leakage Estimation

We consider the problem of measuring how much a system reveals about its secret inputs. We work under the black-box setting: we assume no prior knowledge of the system's internals, and we run the system for choices of secrets and measure its leakage from the respective outputs. Our goal is to estimate the Bayes risk, from which one can derive some of the most popular leakage measures (e.g., min-entropy, additive, and multiplicative leakage). The state-of-the-art method for estimating these leakage measures is the frequentist paradigm, which approximates the system's internals by looking at the frequencies of its inputs and outputs. Unfortunately, this does not scale for systems with large output spaces, where it would require too many input-output examples. Consequently, it also cannot be applied to systems with continuous outputs (e.g., time side channels, network traffic). In this paper, we exploit an analogy between Machine Learning (ML) and black-box leakage estimation to show that the Bayes risk of a system can be estimated by using a class of ML methods: the universally consistent learning rules; these rules can exploit patterns in the input-output examples to improve the estimates' convergence, while retaining formal optimality guarantees. We focus on a set of them, the nearest neighbor rules; we show that they significantly reduce the number of black-box queries required for a precise estimation whenever nearby outputs tend to be produced by the same secret; furthermore, some of them can tackle systems with continuous outputs. We illustrate the applicability of these techniques on both synthetic and real-world data, and we compare them with the state-of-the-art tool, leakiEst, which is based on the frequentist approach.

cs.CR↗

arXiv2020

Optimal Obfuscation Mechanisms via Machine Learning

We consider the problem of obfuscating sensitive information while preserving utility, and we propose a machine learning approach inspired by the generative adversarial networks paradigm. The idea is to set up two nets: the generator, that tries to produce an optimal obfuscation mechanism to protect the data, and the classifier, that tries to de-obfuscate the data. By letting the two nets compete against each other, the mechanism improves its degree of protection, until an equilibrium is reached. We apply our method to the case of location privacy, and we perform experiments on synthetic data and on real data from the Gowalla dataset. We evaluate the privacy of the mechanism not only by its capacity to defeat the classifier, but also in terms of the Bayes error, which represents the strongest possible adversary. We compare the privacy-utility tradeoff of our method to that of the planar Laplace mechanism used in geo-indistinguishability, showing favorable results. Like the Laplace mechanism, our system can be deployed at the user end for protecting his location.

cs.LG↗

arXiv2020

Feature selection in machine learning: Rényi min-entropy vs Shannon entropy

Feature selection, in the context of machine learning, is the process of separating the highly predictive feature from those that might be irrelevant or redundant. Information theory has been recognized as a useful concept for this task, as the prediction power stems from the correlation, i.e., the mutual information, between features and labels. Many algorithms for feature selection in the literature have adopted the Shannon-entropy-based mutual information. In this paper, we explore the possibility of using Rényi min-entropy instead. In particular, we propose an algorithm based on a notion of conditional Rényi min-entropy that has been recently adopted in the field of security and privacy, and which is strictly related to the Bayes error. We prove that in general the two approaches are incomparable, in the sense that we show that we can construct datasets on which the Rényi-based algorithm performs better than the corresponding Shannon-based one, and datasets on which the situation is reversed. In practice, however, when considering datasets of real data, it seems that the Rényi-based algorithm tends to outperform the other one. We have effectuate several experiments on the BASEHOCK, SEMEION, and GISETTE datasets, and in all of them we have indeed observed that the Rényi-based algorithm gives better results.

cs.LG↗

arXiv2019

Full Convergence of the Iterative Bayesian Update and Applications to Mechanisms for Privacy Protection

The iterative Bayesian update (IBU) and the matrix inversion (INV) are the main methods to retrieve the original distribution from noisy data resulting from the application of privacy protection mechanisms. We show that the theoretical foundations of the IBU established in the literature are flawed, as they rely on an assumption that in general is not satisfied in typical real datasets. We then fix the theory of the IBU, by providing a general convergence result for the underlying Expectation-Maximization method. Our framework does not rely on the above assumption, and also covers a more general local privacy model. Finally we evaluate the precision of the IBU on data sanitized with the Geometric, $k$-RR, and RAPPOR mechanisms, and we show that it outperforms INV in the first case, while it is comparable to INV in the other two cases.

cs.CR↗