PTR
A classical differential privacy framework that proposes an upper bound on sensitivity and tests whether a dataset satisfies the condition; if it does, the result is published with noise calibrated to that bound. It is suited to settings with low local sensitivity.
Timeline (last 90 days)
The Propose-Test-Release (PTR) framework proceeds as follows: first propose an upper bound on sensitivity, then test whether the dataset satisfies the condition corresponding to that upper bound (the testing process consumes privacy budget), and if the test passes, release with noise calibrated to that upper bound.
PFR inherits PTR's 'test first, then release' idea, but replaces 'test' with 'filter': it directly constructs a subset satisfying the low-sensitivity condition via privatized filtering, and then releases the result.
All Facts (2)
The Propose-Test-Release (PTR) framework proceeds as follows: first propose an upper bound on sensitivity, then test whether the dataset satisfies the condition corresponding to that upper bound (the testing process consumes privacy budget), and if the test passes, release with noise calibrated to that upper bound.
50%UnverifiedPFR inherits PTR's 'test first, then release' idea, but replaces 'test' with 'filter': it directly constructs a subset satisfying the low-sensitivity condition via privatized filtering, and then releases the result.
50%