Pasin Manurangsi

Rényi differential privacy unlocks tighter privacy guarantees in partition selection, but releasing partition frequencies comes at a cost.

Forget efficient, private $k$-means coresets: this work proves computational lower bounds, showing they're likely impossible to achieve under standard cryptographic assumptions.

Natural privacy filters, despite their promise for tighter privacy accounting, aren't universally "free," limiting their applicability to specific families of differentially private mechanisms.

User-level differential privacy demands a surprising amount of memory: even for basic tasks like counting distinct elements, private algorithms provably require significantly more space than their non-private counterparts.