Synthetic Data within the Risk

1
Synthetic Data within the Risk Utility Framework
Keith Spicer Office for National Statistics
2
Microdata Product Range
AR On Site Access (VML or SDS)
High
PERSONAL INFORMATION Access for Approved
Researchers only
AR Desktop Access

• Disclosure Risk of Dataset

Safeguarded Licensing
Level at which data become Personal Information
NOT PERSONAL INFORMATION
OGL / Public Use
High
Low
Data Utility
3
Microdata Product Range
AR On Site Access (VML or SDS)
High
PERSONAL INFORMATION Access for Approved
Researchers only
AR Desktop Access

• Disclosure Risk of Dataset

Safeguarded Licensing
Level at which data become Personal Information
Target Area for Synthetic Data
NOT PERSONAL INFORMATION
OGL / Public Use
High
Low
Data Utility
4
Utility

• Framework only considers the utility of the data
  in a research context
• Synthesis creates microdata that are not personal
  information (key assumption)
• Ease of Access
• Training
• Testing code (prior to access of Personal
  Information)

5
Synthesised data for research

• Goal to retain research utility while reducing
  disclosure risk
• At what point do real data become synthetic?
• What methods are specifically synthesising
  methods?
• Is synthesising just doing lots of SDC but in a
  smart utility retaining way?