Time is Money: Strategic Timing Games in Proof-of-Stake Protocols

Caspar Schwarz-Schilling¹, Fahad Saleh², Thomas Thiery¹, Jennifer Pan³,

Nihar Shah³, Barnabé Monnot¹

​

¹ Ethereum Foundation, ² Wake Forest University, ³Jump Crypto

​

​

Advances of Financial Technologies, Oct 22nd 2023

1

Safety & Liveness! Economic Fairness?

Consensus protocols are typically evaluated based on their ability to maintain liveness and safety [Cachin, Vukolić, 2019].

Liveness: regular addition of new transactions to the output ledger in a timely manner

Safety: security of confirmed transactions remaining in their positions within the ledger

​

However, blockchain protocols also care about fairness of economic outcomes amongst consensus participants to preserve decentralization.

→ honest protocol participation should be maximally profitable

2

MEV frustrates such fairness goals…

Maximal Extractable Value (MEV) is the value that consensus participants, in their duties as block proposers, accrue by selectively including, excluding and ordering user transactions [Daian et al., 2019].

​

MEV has security implications for consensus protocols, for example:

• Systems predominantly based on tx fee rewards, increased variance in miner rewards may result in consensus instability [Carlsten et al., 2016]
• Similarly, a rational actor issuing a “whale tx” with an abnormally large tx fee can convince peers to fork the current chain [Liao, Katz, 2017]

3

MEV: Time is Money for Consensus Participants

Potential MEV accrues over time as users submit transactions and the value of the set of pending transactions increases for the block proposer.

As a consequence, time is valuable to consensus participants.

This feature is obviated by the assumption of honest behavior in typical models of consensus (honest/byzantine).

Rational consensus participation provides a more accurate portrayal of behavior in economically incentivized systems such as blockchain protocols.

4

Time in Ethereum

5

Progression of a slot

Block proposal “on time”

7

Block proposal “late”

8

Timing Games in Propose-vote Type of PoS Protocols

Proposers

Reward increases linearly with time relative to most recent canonical block so long as their block eventually becomes canonical.

→ Propose block as late as possible, while being early enough to become canonical

Attesters

Correctness: Vote correctly (~beauty contest)

Freshness: Release vote before next proposer releases block

→ Tension between waiting for as long as possible for more MEV/information and being too late to receive any rewards.

9

Static Model

• Time is partitioned into slots with one block proposer and a committee of attesters each round.
• Proposer chooses to extend chain or not and when to release block
• Attesters choose when to release vote for block or vote “empty” (no blocktree)
• Strategies are static: All players choose a schedule of actions at the beginning of the game!

​

10

Equilibrium Analysis

Multiplicity of equilibria: Attesters can coordinate to implement proposers acting at any particular time within the slot!

• Attesters force proposer to adhere to their voting schedule
• Proposer releases block as late as possible given voting schedule

11

Empirical Analysis

12

Proposer-Builder Separation: MEV-Boost

Users submit txs.

Searchers find valuable MEV opportunities and send tx bundles to builders.

Builders build maximally profitable blocks given searcher bundles and public txs. Builders bid for the right to build the block instead of the block proposer.

Relays act as a trust facilitator between builders and proposers by ensuring block validity and bid payment, as well as keeping builder’s block content private.

Proposers can locally build a block or accept a builder bid (via relay) by signing block header.

→ ~90% of all Ethereum validators run MEV-Boost

13

Data

Data points provided by ultra sound relay

• ~185’000 slots (March 4 - April 11 2023)
• ~150m bids
• For ~85’000 slots winning block originated from ultra sound relay
• Collected several timestamps: receivedAt, eligibleAt, signedAt

14

Proposer-Builder Separation: MEV-Boost

15

Marginal Value of Time

16

Dissecting Timing of Block Messages

17

​

receivedAt� Relay receives bid

eligibleAt� Relay validates bid

signedAt� Relay receives signed bid by proposer

Conclusions and Implications

Timing games are rational to engage in and occur, but not yet to their full possible extent.

Good

Attesters can enforce an equilibrium strategy containing timing games.

Bad

Proposers will maximally delay their block given any attestation deadline.

Proposers can delay blocks “better” through sophisticated latency optimizations.

18

Key Takeaways

• Consensus participants are exposed to incentives outside of consensus rewards, namely MEV. Hence, time is valuable to them.
• Block proposers find themselves incentivized to use their temporary monopoly power by engaging in strategic timing games.
• Timing games can be contained but not stopped by committee of attesters.
• Ensuring economic fairness among consensus participants is critical to preserve decentralization.
• Rational consensus participation provides a more accurate portrayal of behavior in economically incentivized systems such as blockchain protocols.

19

More detail in the full paper…

• Formal definition of the game
• Equilibrium proof
• More empirical analysis: investigating proposer types
• …

​

​

​

20

Thank you!

21

Caspar Schwarz-Schilling

Robust Incentives Group (RIG), Ethereum Foundation

caspar@ethereum.org

@casparschwa

A PoS protocol cannot achieve economic fairness amongst rational consensus participants with heterogeneous resources.

​

• Any PoS system needs a mechanism for determining the block proposer.
• This random selection must be done explicitly based on the stake distribution of its bonded consensus participants.
• How to move through this explicit, pseudo-random sequence of leader duties (aka block proposers)?
• Just cycle through it? No! We need the protocol to be live if a proposer goes offline…
• Introducing “deterministic time”: Fixed time intervals (e.g. Ethereum slots) or timeouts (PBFT-style, i.e. Tendermint) that specify for what time interval a proposal is valid. This ensures liveness!
• Either way we end up with a monopoly right for some interval of time.
• Monopoly can be abused as it leaves wiggle room to time block release strategically.
• A sophisticated actor better at abusing monopoly power leading to unequal payoffs amongst consensus participants

22

23

24

25

26