Seminar Details

The ability to securely share or distribute private information among a group of mutually distrusting parties is essential for enabling secure computation on sensitive data. Traditionally, the emphasis has solely been on securely sharing private data. However, recent works have extended these ideas to allow for secure sharing of private functions as well, giving rise to new applications in secure computation.
One particularly useful function in this context is the point function -- a function that outputs zero on all inputs except a specific, pre-defined input, where it outputs a non-zero value. Here, secure sharing means that the point function is split across parties in such a way that each party holds only a portion, or "share", of it. Crucially, these shares are such that authorized subset of them can be combined to reconstruct the function, while other subsets reveal no information about it.This form of secure sharing is enabled by a cryptographic primitive known as a Distributed Point Function (DPF).  DPF allows the point function to be broken into smaller, secure shares, where each share is significantly smaller than the function’s entire input domain. Each party can locally evaluate its share on any input to obtain a share of the output, and combining these yields the correct result. While state-of-the-art DPFs achieve share sizes that are sublinear in the domain size, most of the progress in this area has been limited to the two-party setting. Existing multiparty DPF constructions, however, had share sizes that grew exponentially with the number of parties.

In this talk, I will present the first construction of multiparty DPFs where share sizes remain sublinear in the domain size and grow only polynomially with the number of parties.