Flashbots researcher: Encrypted mempools are not yet economically viable; programmable privacy may be a better path

Foresight News reports that Flashbots Senior Research Scientist and Imperial College London visiting scholar Jonathan Passerat-Palmbach systematically reviewed the academic evolution and real-world limitations of encrypted mempools at the EthCC[9] conference. He pointed out that from early committee-based threshold encryption schemes, to batched threshold encryption, and further to the Beast-Mev scheme combining silent setup with batch decryption, the academic community has proposed various constructions. However, no existing scheme can simultaneously meet all the requirements for protocol-level deployment—such as silent setup, non-interactive decryption, small keys, and small ciphertexts. Therefore, the cryptographic section of encrypted mempools cannot yet be directly embedded into the Ethereum protocol.

The more critical issue, he noted, lies on the economic level. Encrypted mempools use a blind ordering mechanism, where transactions are sorted in an encrypted state. This gives rise to three problems: 1) Searchers will still submit large amounts of blind arbitrage bundles, resulting in on-chain spam transactions; 2) Reduced MEV subsidies will make it harder for geographically remote validators to remain competitive, which will in turn exacerbate centralization; 3) User execution quality will decline, with greater price deviation and higher rollback probability for transactions. Jonathan stated that the sequencing design of decrypting after execution prevents auction mechanisms from functioning effectively, even though auctions are the optimal way to manage MEV incentives. He proposed programmable privacy as an alternative, allowing for intermediate states between fully encrypted and fully public, which can currently be implemented via TEE and could incorporate more advanced solutions such as FHE in the future. In summary, he concluded that privacy cannot exist independently from economic incentives and should continue to be experimented with off-protocol rather than being hastily integrated.