Claim: Statistical noise infusion is a decades-old privacy protection method — TRUE

The claim is that statistical noise infusion — deliberately adding random perturbation to data to prevent identification of individuals — is a decades-old privacy protection method, not a recent invention. That claim is true, and the historical record is unusually well-documented.

The clearest anchor point is Tore Dalenius's 1977 paper 'Towards a Methodology for Statistical Disclosure Control,' published in Statistisk Tidskrift. Dalenius formally introduced the idea that noise or perturbation added to microdata or tabular outputs could shield individual privacy, establishing the intellectual foundation of the entire field. Nine years later, J.J. Kim's 1986 paper in the Proceedings of the American Statistical Association went further, explicitly proposing random noise addition to microdata records as a working disclosure-limitation method. That puts active, named, peer-reviewed proposals for noise infusion at nearly four decades ago — before the World Wide Web existed.

The strongest counterargument is that noise infusion only became practically significant with Cynthia Dwork's 2006 differential privacy framework, and that earlier work was too informal or theoretical to count as real deployment. This is the most honest version of the skeptical case, and it deserves a direct answer. Dwork's paper is genuinely foundational — it provides the rigorous mathematical guarantees that earlier approaches lacked. But Dwork herself builds explicitly on the prior noise-infusion literature, and the technique was already running in production systems before her paper appeared. According to Abowd and Woodcock's 2001 chapter in Confidentiality, Disclosure, and Data Access, the Census Bureau was operationally using noise infusion in its Longitudinal Employer-Household Dynamics program by the late 1990s. A 2002 Census Bureau Statistical Research Division technical report confirms that data perturbation had been used in official statistics for decades prior to that publication.

The institutional record closes the argument. The Federal Committee on Statistical Methodology's Statistical Policy Working Paper 22 — produced under the U.S. Office of Management and Budget — catalogued noise addition as one of several established disclosure-limitation methods in use across federal statistical agencies. Its first edition appeared in 1994; the second in 2005. A method does not get codified in federal statistical policy unless it has already earned operational trust. Conceding the steelman: Dwork's 2006 formalization did represent a genuine scientific advance, and the specific differential privacy implementation the Census Bureau adopted for the 2020 Census is newer and more mathematically precise than what came before. That nuance is real. But it does not change the fact that the underlying technique of noise infusion predates it by roughly three decades.

The manipulation pattern to watch for here runs in both directions. Critics of the 2020 Census differential privacy rollout sometimes imply the method is an untested novelty cooked up by computer scientists with no statistical tradition behind it — that framing is false. Defenders sometimes overstate continuity, glossing over the meaningful differences between 1980s noise-addition heuristics and modern differential privacy guarantees. The honest position is the one the evidence supports: noise infusion is a mature, institutionally validated, genuinely decades-old technique, and its 2020 Census application is a more rigorous evolution of that tradition, not a departure from it.