Learning your identity and disease from research papers: information leaks in genome wide association study

By Rui Wang, Yong Fuga Li, Xiaofeng Wang, Haixu Tang, Xiao-yong Zhou. 2009

262 citations recorded by the source corpus.

Subfields: Cancer Research, Genetics, Privacy & Security, Computational Biology & Bioinformatics

How researchers see this paper

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

In Genetics

In Privacy & Security

In Computational Biology & Bioinformatics

Citations in the corpus

The power of synergy in differential privacy: Combining a small curator with local randomizers (2019, background): “…appears appropriate for it from the privacy perspective – the majority of the data would be analyzed with the guarantee of local di ﬀ erential privacy, and only a small number of data points would be entrusted to a curator, whose analysis’s output should also be di ﬀ erentially private [65].”
Comprehensive Privacy Analysis of Deep Learning: Stand-alone and Federated Learning under Passive and Active White-box Inference Attacks (2018, background): “In a reconstruction attack, the attacker’s objective is to infer attributes of the records in the training set [2], [3].”
Knock Knock, Who's There? Membership Inference on Aggregate Location Data (2017, methodology): “[33] to use correlations within the human genome, reducing the need for prior knowledge about the target.”
Differentially Private Empirical Risk Minimization (2009, neutral): “Moreover, even releasing statistics on a sensitive dataset may not be sufficient to preserve privacy, as illustrated on genetic data (Homer et al., 2008; Wang et al., 2009).”
Comprehensive Privacy Analysis of Deep Learning: Passive and Active White-box Inference Attacks against Centralized and Federated Learning (2018, background): “In a reconstruction attack, the attacker’s objective is to infer attributes of the records in the training set [2], [3].”
Towards Making Systems Forget with Machine Unlearning (2015, background): “Our evaluation, on four diverse learning systems and real-world workloads, shows that our approach is general, effective, fast, and easy to use.”
Machine Learning with Membership Privacy using Adversarial Regularization (2018, background): “The reconstruction attack is a more generic type of attack, where the objective is to infer sensitive attributes of many individuals in the training set [13, 49].”
Differentially private data aggregation with optimal utility (2014, background): “For example, researchers demonstrated how precise information about consumer habits can be reconstructed from the electricity consumption information collected by smart meters [36] and how an individual’s identity and state of health can be derived from genome wide association studies [40].”
Stolen Memories: Leveraging Model Memorization for Calibrated White-Box Membership Inference (2019, background): “, a number of studies [9, 14, 36, 39, 44] have looked into membership attacks on statistics commonly published in genome-wide association studies.”
Machine learning and genomics: precision medicine versus patient privacy (2018, background): “e, such as using genealogies built from public records to identify participants in family-based studies [16], or assessing whether a given genotype is part of a cohort summed up by allele frequencies [17, 18]. As a result, both the NIH and the Wellcome Trust updated their data sharing policy to stro”
Differentially private histogram publication (2012, background): “Furthermore, a strengthened version of this attack [2] threatens the publication of any research paper on genome-wide association studies, which currently is an active field in biomedical research.”
Plausible Deniability for Privacy-Preserving Data Synthesis (2017, background): “A range of inference attacks on, for example, AOL search log dataset [2], Netﬂix movie rating dataset [39], Genomic data [48, 22], location data [18, 46], and social networks data [40], shows that simple modiﬁcation of sensitive data by removing identiﬁers or by generalizing/suppressing data…”
Countering GATTACA: efficient and secure testing of fully-sequenced human genomes (2011, background): “LATED WORK As discussed in Section 1, traditional approaches to privacy, such as de-identiﬁcation, are often ineﬀective on genomic data, since the genome itself is the ultimate identiﬁer. We refer to [57, 41, 79, 86] for details on privacy risks associated to releasing genomic information, even when”
Membership Privacy in MicroRNA-based Studies (2016, background): “Wang et al. extend this attack by making use of the correlations among the different positions in the genome [34].”
Privacy Risk in Machine Learning: Analyzing the Connection to Overfitting (2017, result): “A number of previous studies [21], [22], [47], [48], [49], [50] have looked into membership attacks from statistics commonly published in genome-wide association studies (GWAS).”
Differentially Private Data Generative Models (2018, background): “This attack was then extended by Wang et al. [63] by performing correlation attacks, without prior knowledge about the target.”