Simson L. Garfinkel

Simson L. Garfinkel, Philip Leclerc

The U.S. Census Bureau is using differential privacy (DP) to protect confidential respondent data collected for the 2020 Decennial Census of Population & Housing. The Census Bureau's DP system is implemented in the Disclosure Avoidance System (DAS) and requires a source of random numbers. We estimate that the 2020 Census will require roughly 90TB of random bytes to protect the person and household tables. Although there are critical differences between cryptography and DP, they have similar requirements for randomness. We review the history of random number generation on deterministic computers, including von Neumann's "middle-square" method, Mersenne Twister (MT19937) (previously the default NumPy random number generator, which we conclude is unacceptable for use in production privacy-preserving systems), and the Linux /dev/urandom device. We also review hardware random number generator schemes, including the use of so-called "Lava Lamps" and the Intel Secure Key RDRAND instruction. We finally present our plan for generating random bits in the Amazon Web Services (AWS) environment using AES-CTR-DRBG seeded by mixing bits from /dev/urandom and the Intel Secure Key RDSEED instruction, a compromise of our desire to rely on a trusted hardware implementation, the unease of our external reviewers in trusting a hardware-only implementation, and the need to generate so many random bits.

Simson L. Garfinkel, John M. Abowd, Sarah Powazek

When differential privacy was created more than a decade ago, the motivating example was statistics published by an official statistics agency. In attempting to transition differential privacy from the academy to practice, the U.S. Census Bureau has encountered many challenges unanticipated by differential privacy's creators. These challenges include obtaining qualified personnel and a suitable computing environment, the difficulty accounting for all uses of the confidential data, the lack of release mechanisms that align with the needs of data users, the expectation on the part of data users that they will have access to micro-data, and the difficulty in setting the value of the privacy-loss parameter, $ε$ (epsilon), and the lack of tools and trained individuals to verify the correctness of differential privacy implementations.