Answers are grounded in this paper and cite their sources.
Differential privacy provides the first theoretical foundation with provable privacy guarantee against adversaries with arbitrary prior knowledge. The main idea to achieve differential privacy is to inject random noise into statistical query results. Besides correctness, the most important goal in the design of a differentially private mechanism is to reduce the effect of random noise, ensuring that the noisy results can still be useful.
No takes yet. Share an insight, caveat, or question.
Li et al. (2011) studied this question.