OGST - Ewa Syta
Title: Securing Distributed Randomness from Active Bias Attacks
Abstract: Distributed protocols among unreliable, mutually distrustful parties often depend for security on commonly agreed-upon public random values. It is hard to prevent malicious participants from biasing this randomness, however, by manipulating their explicit inputs or by selectively injecting failures. We explore this challenge by analyzing different ways to introduce bias as well as propose a bias-resistant randomness protocol that produces a common public random value. Clients can either use that value directly as needed or supply it to produce lottery tickets to choose a subset from a set in a bias-resistant fashion.
The protocol we propose uses threshold cryptography in a small group to agree on the common bias-resistant value. The main ideas to achieve unbiasiability are to remove the incentives for dishonest users to misbehave by forcing them to fully commit to their input and actions before learning anything about the protocol output, and to allow honest members to carry out the protocol on their own, assuming that the protocol reaches a certain point, that we term “the barrier”.
Additionally, we explore ways to achieve scalability and client church tolerance while preserving bias-resistance.