ML05: Protecting Data for Official Statistics with Formal Privacy

Music City Center 

Methodology for disclosure avoidance (DA) is critically important for survey methodologists and governments in producing official statistics. Indeed, nearly all surveys are conducted under a pledge to maintain the confidentiality of respondents' sensitive information. The choice of DA method significantly impacts data quality and determines the types of analyses that can be conducted.
Formally private methods, which provide mathematically provable privacy guarantees, have recently gained significant attention as the risk of disclosure breaches increases. Advancements in sophisticated tools, such as AI and machine learning, have made data more vulnerable to attacks. At the same time, the demand for more granular and refined data continues to grow, driven by these same technological advancements. While straightforward to implement, the use of formally private method can severely compromise data utility if not applied creatively and thoughtfully.
During this roundtable, we will discuss issues and challenges around implementing formally private methods to data used to produce official statistics as well as current projects in the government to transition to these disclosure limitati

Disclosure Limitation

Establishment Data

Differential Privacy

Survey Data

Data Protection 

Government Statistics Section