Glossary
Differential Privacy
Differential privacy adds carefully calibrated statistical noise to the results of queries over a dataset, so that whether or not any single individual's record is included barely changes the output. Unlike suppression-based techniques, it provides this guarantee even against someone running many different queries and trying to triangulate an individual's data by comparing the results.
The calibration is a genuine tradeoff, not a free guarantee: more noise gives a stronger privacy guarantee and a less precise answer; less noise gives a more useful answer with a weaker guarantee. Getting this balance right for a specific use case requires deliberately choosing where on that tradeoff to sit, not just enabling the technique and assuming a single default setting is appropriate everywhere it's applied.