Utility-Privacy Tradeoffs in Databases: An Information-Theoretic Approach

By Lalitha Sankar, S. Rajagopalan, H. Vincent Poor. 2011

390 citations recorded by the source corpus.

Ensuring the usefulness of electronic data sources while providing necessary privacy guarantees is an important unsolved problem. This problem drives the need for an analytical framework that can quantify the privacy of personally identifiable information while still providing a quantifiable benefit (utility) to multiple legitimate information consumers. This paper presents an information-theoretic framework that promises an analytical model guaranteeing tight bounds of how much utility is possible for a given level of privacy and vice-versa. Specific contributions include: 1) stochastic data models for both categorical and numerical data; 2) utility-privacy tradeoff regions and the encoding (sanization) schemes achieving them for both classes and their practical relevance; and 3) modeling of prior knowledge at the user and/or data source and optimal encoding schemes for both cases.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Extremal Mechanisms for Local Differential Privacy (2014, background): “The Kullback-Leibler (KL) divergence Dkl(M0||M1) is a special case of f -divergence with f(x) = x log x, and total variation ‖M0 −M1‖TV is a special case with f(x) = (1/2)|x − 1|.”
Mobile sensor data anonymization (2018, background): “Another approach to privacy-preserving data release is based on information theory [17, 28], where the mutual information, between the released data and the latent information which can be drawn from data, is considered as the measure of privacy.”
Shredder: Learning Noise Distributions to Protect Inference Privacy (2019, methodology): “dely used in the literature for both understanding the behaviour of neural networks [24, 25, 24, 26, 27, 28], and also to quantify information leakage in anonymity systems in the context of databases [29, 30, 31]. We also use the reverse of mutual information (1=MI) as our main and ﬁnal notion of pr”
Adversarially Learned Representations for Information Obfuscation and Inference (2019, background): “There are theoretical works on studying utility-privacy tradeoffs using information metrics (Sankar et al., 2013; Basciftci et al., 2016), but, to our knowledge, none have been applied to highdimensional data (e.”
On the Robustness of Information-Theoretic Privacy Measures and Mechanisms (2018, methodology): “, Shannon’s mutual information [9, 10], k-anonymity [11], differential privacy [12], maximal leakage [13], total variation [14], among others.”
Generative Adversarial Privacy: A Data-Driven Approach to Information-Theoretic Privacy (2018, background): “Context-aware privacy notions fall under the domain of information theoretic (IT) privacy; see, for example, [4, 5, 25, 26, 27].”
Differentially Private Testing of Identity and Closeness of Discrete Distributions (2017, background): “Information theoretic approaches to data privacy have been studied recently using quantities like mutual information, and guessing probability to quantify privacy [46, 52, 21, 59, 39].”
Learning Generative Adversarial RePresentations (GAP) under Fairness and Censoring Constraints (2019, background): “, mutual information censoring (Rebollo-Monedero et al., 2010; Calmon and Fawaz, 2012; Sankar et al., 2013; Salamatian et al., 2015; Basciftci et al., 2016), achieve improved censoring-utility tradeoff, but assume that the data holder has access to dataset statistics.”
Obfuscation via Information Density Estimation (2019, background): “1 Related Work The problem of balancing the competing objectives of providing meaningful information and inference from disclosed data, on the one hand, and obfuscating sensitive information, on the other hand, has been widely studied in information-theoretic privacy [5–9, 34, 35].”
A rate-disortion perspective on local differential privacy (2014, background): “In this paper we will examine the relationship between two of these: information-theoretic privacy [3] and differential privacy [4].”
Technical Privacy Metrics (2015, background): “In this case, normalized mutual information measures the number of bits leaked on average from any entry [109].”
Shredder: Learning Noise to Protect Privacy with Partial DNN Inference on the Edge (2019, background): “Mutual information has been widely used in the literature for both understanding the behaviour of neural networks [24, 25, 24, 26, 27, 28], and also to quantify information leakage in anonymity systems in the context of databases [29, 30, 31].”
On the relation between identifiability, differential privacy, and mutual-information privacy (2014, neutral): “In this paper, we consider a form of identifiability that prevents the input database from being identified given the observation of the output database, prior knowledge and the mechanism.”