PDF - Synthetic Data by a Generative Adversarial Network - ON-184

Preferred Disciplines: Computer Science, Engineering, Applied Mathematics/Statistics (Post-Doc)
Company: Privacy Analytics Inc (an IQVIA company)
Project Length: 3 years (9 units)
Desired start date: January 2019
Location: Ottawa, ON
No. of Positions: 1
Preferences: None

About the Company: 

A global authority in data anoymization, providing training, professional services, and software products. 

Project Description:

Data can be generated synthetically from source data in a variety of ways, as a mean of providing safe data for public use. However, even fully synthetic data can be subject to potential disclosure risks depending on how “close” they represent the source data. There has been some very recent work exploring the use of a generative adversarial network (GAN) to create synthetic data. We wish to develop this technology further to create realistic and complex data from any structured source, incorporating privacy measures from the fields of statistical disclosure control and computer science.

Research Objectives:

  • Determine or develop the generative adversarial network that will allow for the creation of synthetic data for complex structures (e.g., longitudinal)
  • Incorporate measures of disclosure risk into a generative adversarial network
  • Build the appropriate tools and techniques for the sharing and release of synthetic data from a generative adversarial network

Methodology:

  • Generative adversarial networks
  • Syntactic as well as proveable measures of data privacy
  • Statistical disclosure control methods used to produce safe data 

Expertise and Skills Needed:

  • Thourough knowledge and understanding of generative adversarial networks, with a preference for candidates that have implemented solutions
  • Knowledge of, or willingness to learn, methods and approaches of statistical disclosure control and data privacy
  • Knowledge and understanding of applied statistics and computer science
  • Practical and solutions driven for real-world application of data privacy
  • Strong programming skills for data analysis and for implementing a generative adversarial network (e.g., R, Python).

For more info or to apply to this applied research position, please

  1. Check your eligibility and find more information about open projects
  2. Interested students need to get the approval from their supervisor and send their CV along with a link to their supervisor’s university webpage to Mel Chaar
Program: