TrX: Encrypted Mempools in High Performance BFT Protocols
Session summary
The presenter describes an encrypted mempool design for a BFT blockchain that aims to reduce harmful MEV by hiding transaction contents from block proposers while preserving the network’s existing trust model. Users encrypt transactions client‑side to a public key whose secret is shared among validators, so the mempool holds ciphertext. The speaker explains that a straightforward threshold decryption per transaction is prohibitively expensive, causing quadratic communication and computation and adding latency on the consensus critical path. This motivates using batch threshold encryption (BTE), which operates once per block on a compact digest and yields a decryption key that unlocks exactly that block’s transactions.
The authors propose a new BTE construction that avoids encrypting to a fixed round or slot and uses a BLS‑style key pair without message hashing, combined with randomization from a common reference string (CRS). This enables standard per‑epoch DKG while shifting a large CRS into a one‑time trusted setup. They report an implementation integrated into Aptos consensus, with pipelined reconstruction moving most work off the critical path. In experiments on 50 geo‑simulated nodes at 1,000 TPS, they observe modest additional latency compared to baseline and demonstrate a testnet wallet flow for optional encrypted submissions.
Key takeaways
- The authors seek to mitigate harmful MEV by making the mempool private via client‑side encryption to a validator‑shared key, without adding trusted leaders, hardware, or off‑chain committees.
- Naive per‑transaction threshold decryption leads to quadratic overhead and significant on‑path latency, so the authors instead use batch threshold encryption over a block digest with a single reconstruction pipeline.
- Their scheme employs a BLS‑like key pair plus a CRS‑based randomization, supports efficient per‑epoch DKG, and in an Aptos prototype shows low additional latency when combined with pipelined reconstruction and optional encrypted transactions.