In 2018 the Census Bureau discovered that results of the 2010 census could be processed and matched with external sources in such a way as to reveal confidential personal information, in violation of the law.
“This had not been thought to be feasible owing to the large amount of data and computation involved,” a new report from the JASON science advisory panel said. But in fact it was feasible, the JASONs confirmed. The risk of re-identifying protected personal information “is four orders of magnitude larger than had been previously assessed.”
In order to prevent this potential privacy violation, the Census Bureau proposes to use an approach called Differential Privacy. This method, developed by Cynthia Dwork and colleagues, adds “tailored noise” to the results of any query of the census data. Doing so makes it possible “to publish information about a survey while limiting the possibility of disclosure of detailed private information about survey participants.”
The JASONs affirmed that the Differential Privacy technique would provide the necessary privacy protection but said that it would come with a cost in accuracy, particularly with respect to small data groups.
“As the size of the population under consideration becomes smaller, the contributions from injected noise will more strongly affect such queries. Note that this is precisely what one wants for confidentiality protection, but is not desirable for computation of statistics for small populations.”
See Formal Privacy Methods for the 2020 Census, JASON report JSR-19-2F, April 2020.
“Depending on the ultimate level of privacy protection that is applied for the 2020 census, some stakeholders may well need access to more accurate data,” the JASONs said.
“A benefit of differential privacy is that products can be released at various levels of protection depending on the [desired] level of statistical accuracy. The privacy-loss parameter can be viewed as a type of adjustable knob by which higher settings lead to less protection but more accuracy. However, products publicly released with too low a level of protection will again raise the risk of re-identification,” the new report said.
* * *
The JASONs are currently working on a new study of how to safely reopen research laboratories during the ongoing pandemic, Jeffrey Mervis of Science Magazine reported this week.
“Jason is examining such issues as the 1.8-meter separation rule, the efficacy of personal protective equipment, and the optimal way to reconfigure work space given how the virus is thought to spread.” See “Secretive Jasons to offer advice on how to reopen academic labs shut by pandemic,” May 11.
Called today to speak on behalf of U.S. science and technology, Dr. Jedidah Isler, astrophysicist, educator, strategist, policy-maker, and science communicator, will provide constructive, nonpartisan feedback to the House Committee’s hearing “American Global Competitiveness at 250: Legislative Proposals to Secure U.S. Technology Leadership.”
“Federal data and access to it is not a partisan issue. It is a people issue. Our country cannot achieve greatness without access to the data that measure what we value, who we are, and where we’re heading.”
The United States’ biosecurity governance system is structurally incapable of detecting and responding to certain classes of threats. U.S. biosecurity tools have not kept pace with technological advancements or a changing threat landscape.
The United States has never lacked for scientific ambition. What we need now is a renewed civic commitment to ensuring that talent is harnessed for the benefit of all people. Science can work for everyone. Join us as we build a broader coalition committed to that vision.