publication . Article . Preprint . Other literature type . 2019

Differentially Private Confidence Intervals for Empirical Risk Minimization

Wang, Yue; Kifer, Daniel; Lee, Jaewoo;
Open Access
  • Published: 30 Mar 2019 Journal: Journal of Privacy and Confidentiality, volume 9 (eissn: 2575-8527, Copyright policy)
  • Publisher: Journal of Privacy and Confidentiality
Abstract
<jats:p>The process of data mining with differential privacy produces results that are affected by two types of noise: sampling noise due to data collection and privacy noise that is designed to prevent the reconstruction of sensitive information. In this paper, we consider the problem of designing confidence intervals for the parameters of a variety of differentially private machine learning models. The algorithms can provide confidence intervals that satisfy differential privacy (as well as the more recently proposed concentrated differential privacy) and can be used with existing differentially private mechanisms that train models using objective perturbation...
Subjects
free text keywords: Data mining, computer.software_genre, computer, Empirical risk minimization, Differential privacy, Information sensitivity, Sampling (statistics), Data collection, Confidence interval, Computer science, Objective Perturbation, Output Perturbation, Confidence Intervals, Technology, T, Social Sciences, H, Computer Science - Learning, Computer Science - Cryptography and Security, Statistics - Machine Learning
Related Organizations
47 references, page 1 of 4

[1] Ipums-usa, 2017.

[2] Minnesota population center. integrated public use microdata series, international: Version 6.5 brazil, 2017.

[3] J. Acharya, Z. Sun, and H. Zhang. Di erentially private testing of identity and closeness of discrete distributions. arXiv preprint arXiv:1707.05128, 2017.

[4] A. F. Barrientos, J. P. Reiter, A. Machanavajjhala, and Y. Chen. Di erentially private signi cance tests for regression coe cients. arXiv preprint arXiv:1705.09561, 2017.

[5] R. Bassily, A. Smith, and A. Thakurta. Private empirical risk minimization: E cient algorithms and tight error bounds. In Foundations of Computer Science (FOCS), 2014 IEEE 55th Annual Symposium on, pages 464{473. IEEE, 2014.

[6] A. Blum, C. Dwork, F. McSherry, and K. Nissim. Practical privacy: the sulq framework. In Proceedings of the twenty-fourth ACM SIGMOD-SIGACT-SIGART symposium on Principles of database systems, pages 128{138. ACM, 2005.

[7] M. Bun and T. Steinke. Concentrated di erential privacy: Simpli cations, extensions, and lower bounds. In Theory of Cryptography Conference, pages 635{658. Springer, 2016.

[8] B. Cai, C. Daskalakis, and G. Kamath. Privit: Private and sample e cient identity testing. In International Conference on Machine Learning, pages 635{644, 2017. [OpenAIRE]

[9] O. Chapelle. Training a support vector machine in the primal. Neural computation, 19(5):1155{1178, 2007.

[10] K. Chaudhuri, C. Monteleoni, and A. D. Sarwate. Di erentially private empirical risk minimization. Journal of Machine Learning Research, 12(Mar):1069{1109, 2011. [OpenAIRE]

[11] K. Chaudhuri, A. Sarwate, and K. Sinha. Near-optimal di erentially private principal components. In Advances in Neural Information Processing Systems, pages 989{997, 2012. [OpenAIRE]

[12] Y. Chen, A. Machanavajjhala, J. P. Reiter, and A. F. Barrientos. Di erentially private regression diagnostics. In Data Mining (ICDM), 2016 IEEE 16th International Conference on, pages 81{90. IEEE, 2016. [OpenAIRE]

[13] V. D'Orazio, J. Honaker, and G. King. Di erential privacy for social science inference. 2015. [OpenAIRE]

[14] C. Dwork, K. Kenthapadi, F. McSherry, I. Mironov, and M. Naor. Our data, ourselves: Privacy via distributed noise generation. In EUROCRYPT, 2006. [OpenAIRE]

[15] C. Dwork, F. McSherry, K. Nissim, and A. Smith. Calibrating noise to sensitivity in private data analysis. In Theory of Cryptography Conference, pages 265{284. Springer, 2006.

47 references, page 1 of 4
Abstract
<jats:p>The process of data mining with differential privacy produces results that are affected by two types of noise: sampling noise due to data collection and privacy noise that is designed to prevent the reconstruction of sensitive information. In this paper, we consider the problem of designing confidence intervals for the parameters of a variety of differentially private machine learning models. The algorithms can provide confidence intervals that satisfy differential privacy (as well as the more recently proposed concentrated differential privacy) and can be used with existing differentially private mechanisms that train models using objective perturbation...
Subjects
free text keywords: Data mining, computer.software_genre, computer, Empirical risk minimization, Differential privacy, Information sensitivity, Sampling (statistics), Data collection, Confidence interval, Computer science, Objective Perturbation, Output Perturbation, Confidence Intervals, Technology, T, Social Sciences, H, Computer Science - Learning, Computer Science - Cryptography and Security, Statistics - Machine Learning
Related Organizations
47 references, page 1 of 4

[1] Ipums-usa, 2017.

[2] Minnesota population center. integrated public use microdata series, international: Version 6.5 brazil, 2017.

[3] J. Acharya, Z. Sun, and H. Zhang. Di erentially private testing of identity and closeness of discrete distributions. arXiv preprint arXiv:1707.05128, 2017.

[4] A. F. Barrientos, J. P. Reiter, A. Machanavajjhala, and Y. Chen. Di erentially private signi cance tests for regression coe cients. arXiv preprint arXiv:1705.09561, 2017.

[5] R. Bassily, A. Smith, and A. Thakurta. Private empirical risk minimization: E cient algorithms and tight error bounds. In Foundations of Computer Science (FOCS), 2014 IEEE 55th Annual Symposium on, pages 464{473. IEEE, 2014.

[6] A. Blum, C. Dwork, F. McSherry, and K. Nissim. Practical privacy: the sulq framework. In Proceedings of the twenty-fourth ACM SIGMOD-SIGACT-SIGART symposium on Principles of database systems, pages 128{138. ACM, 2005.

[7] M. Bun and T. Steinke. Concentrated di erential privacy: Simpli cations, extensions, and lower bounds. In Theory of Cryptography Conference, pages 635{658. Springer, 2016.

[8] B. Cai, C. Daskalakis, and G. Kamath. Privit: Private and sample e cient identity testing. In International Conference on Machine Learning, pages 635{644, 2017. [OpenAIRE]

[9] O. Chapelle. Training a support vector machine in the primal. Neural computation, 19(5):1155{1178, 2007.

[10] K. Chaudhuri, C. Monteleoni, and A. D. Sarwate. Di erentially private empirical risk minimization. Journal of Machine Learning Research, 12(Mar):1069{1109, 2011. [OpenAIRE]

[11] K. Chaudhuri, A. Sarwate, and K. Sinha. Near-optimal di erentially private principal components. In Advances in Neural Information Processing Systems, pages 989{997, 2012. [OpenAIRE]

[12] Y. Chen, A. Machanavajjhala, J. P. Reiter, and A. F. Barrientos. Di erentially private regression diagnostics. In Data Mining (ICDM), 2016 IEEE 16th International Conference on, pages 81{90. IEEE, 2016. [OpenAIRE]

[13] V. D'Orazio, J. Honaker, and G. King. Di erential privacy for social science inference. 2015. [OpenAIRE]

[14] C. Dwork, K. Kenthapadi, F. McSherry, I. Mironov, and M. Naor. Our data, ourselves: Privacy via distributed noise generation. In EUROCRYPT, 2006. [OpenAIRE]

[15] C. Dwork, F. McSherry, K. Nissim, and A. Smith. Calibrating noise to sensitivity in private data analysis. In Theory of Cryptography Conference, pages 265{284. Springer, 2006.

47 references, page 1 of 4
Powered by OpenAIRE Open Research Graph
Any information missing or wrong?Report an Issue
publication . Article . Preprint . Other literature type . 2019

Differentially Private Confidence Intervals for Empirical Risk Minimization

Wang, Yue; Kifer, Daniel; Lee, Jaewoo;