Differential privacy implies that adding or deleting a single record does not significantly affect the result of any analysis.
You can find futher information about Differential Privacy Models here
