Distributed Randomness Using Weighted VUFs

Shared randomness in blockchain can expand its support for randomized applications and can also help strengthen its security. Many existing blockchains rely on external randomness beacons for shared randomness, but this approach reduces fault tolerance, increases latency, and complicates application development. An alternate approach is to let the blockchain validators generate fresh shared randomness themselves once for every block. We refer to such a design as the on-chain randomness. In this paper, we design an efficient on-chain randomness protocol for Byzantine fault-tolerance based Proof-of-Stake blockchains with weighted validators. A key component of our protocol is a weighted verifiable unpredictable function (VUF). The notable feature of our weighted VUF is that the computation and communication costs of parties are independent of their weight. This is crucial for scalability of on-chain randomness where we repeatedly evaluate the weighted VUF in quick succession. We also design a new scalable publicly verifiable secret sharing (PVSS) scheme with aggregatable transcript and use it to design a distributed key generation (DKG) protocol for our VUF. We implemented our schemes on top of Aptos, a proof-of-stake blockchain deployed in production, conducted an end-to-end evaluation with 112 validators and a total weight of up to 4053. In this setup, our on-chain randomness protocol adds only 133 milliseconds of latency compared to a protocol without randomness. We also demonstrate the performance improvements of our design through rigorous comparison with baseline methods.