User Tools

Site Tools


6a.052.UVA - Privacy Preserving Multi-Party Analytics

Project - Summary

Preserving privacy in machine learning on multi-party data is of importance to many domains. In practice, existing solutions suffer from several critical limitations, such as significantly reduced utility under privacy constraints or excessive communication burden between the information fusion center and local data providers. In this project, we propose and implement a new distributed deep learning framework that addresses these shortcomings and preserves privacy more efficiently than previous methods. During the stochastic gradient descent training of a deep neural network, we focus on the parameters with large absolute gradients in order to save privacy budget consumption. We adopt a generalization of the Report-Noisy-Max algorithm in differential privacy to select these gradients and prove its privacy guarantee rigorously. Inspired by the recent novel idea of Terngrad [1], we also quantize the released gradients to ternary levels {-B, 0, B}, where B is the bound of gradient clipping. Applying Terngrad can significantly reduce the communication cost without incurring severe accuracy loss. Furthermore, we evaluate the performance of our method on a real-world credit card fraud detection data set consisting of millions of transactions.

Project - Team

Team Member Role Email Phone Number Academic Site/IAB
Steven Boker PI (434) 243-7275 University of Virginia
Yang Wang Graduate Student (434) 466-8381 University of Virginia
Steven Greenspan Project Mentor Not available Not available CA Technologies

Project - Deliverables

1 Research paper on the algorithms and theories of distributed privacy preserving deep learning model.
2 Trained deep learning models on real-world credit card fraud dataset.
3 Empirical evaluation on the utility-privacy trade-off of the multiparty privacy-preserving framework.

Project - Presentation Video

Project - Documents

FilenameFilesizeLast modified
6a.052.uva_executive_summary.docx50.0 KiB2019/08/14 15:37
6a.052.uva_poster_pdf.pdf337.5 KiB2019/08/14 15:37
6a.052.uva_poster_ppt.pptx80.0 KiB2019/08/14 15:37
6a.052.uva_confluence_project_page.pdf149.1 KiB2019/08/14 15:37
6a.052.uva_cvdi_mid-year_report.pdf92.5 KiB2019/08/14 15:37
6a.052.uva_poster_2018_spring_meeting.pdf413.6 KiB2019/08/14 15:37
6a.052.uva_final_project_report.pdf423.2 KiB2019/08/14 15:37
6a.052.uva_quadchart_ppt.pptx55.7 KiB2019/08/14 15:37
6a.052.uva_privacy_preserving_multi-party_analytics_poster_2017_fall_meeting.pptx118.7 KiB2019/08/14 15:37
projects/year6/6a.052.uva.txt · Last modified: 2021/06/02 17:13 by sally.johnson