Researchers created efficient algorithms that estimate how often each outcome occurs in a group while protecting every person's whole batch of data. The methods need essentially the least data theory allows — a mark no earlier efficient method had reached.
When organizations learn how often things happen across many people, privacy rules usually mean collecting extra data. This work pins down that price for a core statistics task and proves a floor no efficient method can beat — privacy's cost becomes a known, closed question.
✦ Protecting every person no longer has to blur the pattern.