Genome-wide association studies (GWASs) have received increasing attention to understand how genetic variation affects different human traits. In this paper, we study whether and to what extent exploiting the GWAS statistics can be used for inferring private information about a human individual. We first provide a method to construct a three-layered Bayesian network explicitly revealing the conditional dependency between single-nucleotide polymorphisms (SNPs) and traits from public GWAS catalog. The key challenge in building a Bayesian network from GWAS statistics is the specification of the conditional probability table of a variable with multiple parent variables. We employ the models of independence of causal influences which assume that the causal mechanism of each parent variable is mutually independent. We then formulate three inference problems based on the dependency relationship captured in the Bayesian network, namely trait inference given SNP genotype, genotype inference given trait, and trait inference given known traits, and develop efficient formulas and algorithms. Different from previous work, the possible target of these inference problems we study may be any individual, not limited to GWAS participants. Empirical evaluations show the effectiveness of our proposed methods. In summary, our work implies that meaningful information can be inferred from modeling GWAS statistics, and appropriate privacy protection mechanisms need to be developed to protect genetic privacy not only of GWAS participants but also regular individuals.

In recent years, deep learning has spread beyond both academia and industry with many exciting real-world applications. The development of deep learning has presented obvious privacy issues. However, there has been lack of scientific study about privacy preservation in deep learning. In this paper, we concentrate on the auto-encoder, a fundamental component in deep learning, and propose the deep private auto-encoder (dPA). Our main idea is to enforce ε-differential privacy by perturbing the objective functions of the traditional deep auto-encoder, rather than its results. We apply the dPA to human behavior prediction in a health social network. Theoretical analysis and thorough experimental evaluations show that the dPA is highly effective and efficient, and it significantly outperforms existing solutions.

Data privacy in genome-wide association studies (GWAS) is a critical yet under-exploited research area. In this paper, we first provide a method to construct a two-layered bayesian network explicitly revealing the conditional dependency between SNPs and traits, from the public GWAS catalog. Then we develop efficient algorithms for two attacks: identity inference attack and trait inference attack based on reasoning with the dependency relationship captured in the constructed bayesian network. Different from previously proposed attacks, the possible target of our attacks may be any common people, not limited to GWAS participants. The empirical evaluations show that unprotected statistics released from GWAS can be exploited by attackers to identify individual or derive private information. Thus we show that mining GWAS statistics threatens the privacy of a much wider population and privacy protection mechanisms should be employed.

Enabling accurate analysis of social network data while preserving differential privacy has been challenging since graph features such as clustering coefficient or modularity often have high sensitivity, which is different from traditional aggregate functions (e.g., count and sum) on tabular data. In this paper, we treat a graph statistics as a function f and develop a divide and conquer approach to enforce differential privacy. The basic procedure of this approach is to first decompose the target computation f into several less complex unit computations f1, · · · , fmconnected by basic mathematical operations (e.g., addition, subtraction, multiplication, division), then perturb the output of each fiwith Laplace noise derived from its own sensitivity value and the distributed privacy threshold ϵi, and finally combine those perturbed fias the perturbed output of computation f. We examine how various operations affect the accuracy of complex computations. When unit computations have large global sensitivity values, we enforce the differential privacy by calibrating noise based on the smooth sensitivity, rather than the global sensitivity. By doing this, we achieve the strict differential privacy guarantee with smaller magnitude noise. We illustrate our approach by using clustering coefficient, which is a popular statistics used in social network analysis. Empirical evaluations show the developed divide and conquer approach outperforms the direct approach.