TLDR:
- Bitcoin miners currently receive two incentives for honest mining: block rewards and transaction fees.
- Using a Bitcoin mining simulation, different mining strategies are used to model potential outcomes of a transaction-fee only mining incentive.
- Under a transaction-fee-only incentive design it advantageous for miners to fork “wealthy” blocks. Selfish mining becomes easier for miners with relatively low hash power. The result is a decrease in overall chain security and continuity.
Core Research Question
If the only expected reward for Bitcoin mining is a transaction fee, will miners be incentivized enough to continue honest mining practices?
Citation
Background
There are currently two incentives for mining Bitcoin: block rewards and transaction fees. For every 210,000 blocks mined, the number of Bitcoins awarded per block mined halves. By 2140, the expected block reward will be zero, and the only monetary incentive miners will have to continue honestly mining will be transaction fees.
Background Terms
- Block reward: the amount of a cryptocurrency received for successfully mining a block. For this research summary, it specifically refers to Bitcoin.
- Transaction fee: a fee paid by users to have their transaction(s) verified.
- Miners: actors who receive compensation for computing power spent (block rewards and transaction fees). They signal valid transactions by including them in the blocks that they mine and broadcast.
- Undercutting attacks: when a miner intentionally forks the longest chain to maximize their reward.
- Selfish mining: a subversive strategy in proof-of-work (PoW) blockchain systems where malicious miners do not release newly mined blocks. This causes good faith miners to use computing power to continue mining blocks already completed by the malicious miner, resulting in the malicious miner focusing on unmined blocks to increase their chances of completing the next block.
- Petty compliance: when a miner chooses to fork the longest chain by picking the block that results in the largest monetary payout instead of the oldest block.
- Default compliance: when a miner chooses to mine the oldest block to continue the longest chain, regardless of payout.
- Mining gap: if the only expected mining payout is a transaction fee, there may be a time when there are no waiting transactions when payout is zero. This can occur immediately after a block is mined. Given there is a non-zero electricity cost for mining at this time, the expected payout would be negative. This may result in miners mining for smaller fractions of time between block arrivals, which may drop the effective hash rate.
Summary
- The researchers provide an overview of mining incentives and potential threats to honest mining.
- They specifically discuss Eyal and Sirer’s deviant mining strategy and Sapirshtein et al’s work analyzing selfish mining.
- The researchers define the assumptions of a Bitcoin mining simulation model they use as their basis for their research, and define a formal language for modeled mining strategies.
- An overview of the mining strategy simulator created by the authors is provided.
- Potential mining strategies, the processes of the simulator, configurable parameters, and the limitations of the simulator are discussed.
- Researchers define potential new deviant mining behavior by testing different parameters using the mining simulator.
- The researchers then discuss and support findings indicating selfish mining strategies are more profitable under the transaction fee model.
- They discuss the limitations of their work and propose future research regarding different monetary incentive designs.
Method
- The researchers use a Bitcoin mining strategy simulator to uncover the potential effects of only relying on the “time-varying nature of transaction fees.” (pg 3)
- By simplifying the assumptions used in the model, researchers find more robust outcomes as the effects should become compounded under more advanced/complicated conditions.
- The Bitcoin mining strategy simulator is described below:
- The base assumptions of the model are as follows:
- The block reward = 0
- Transaction fees are the only source of revenue for miners
- Transactions arrive at a constant rate
- The network has zero latency
- Miners can choose transaction fees
- Miners can always include all available transactions
- The simulator is able to use the following parameters to investigate different scenarios:
- T = total transaction fees (and total transactions)
- B = block
- Tx(B) = set of transactions (Tx) included in the block (B)
- Rem(B) = remaining (Rem) transactions (T) after block (B). These are transactions not included in block (B) nor in any block before block (B).
- Height(B) = height of a chain ending at block (B).
- Owner(B) = the miner (Owner) that produced block (B).
- The simulator is time-driven, allowing for new mining strategies to be introduced.
- Each round is considered one unit of time, in each round miners decide which block to extend and which transactions to include.
- A “game” is set using the above parameters and allowed to run for several rounds. A record of rewards earned by the miners in each game is kept.
- Simulations can consist of a single or multiple games.
- The simulator utilizes a no-regret learning methodology ensuring the results converge to a coarse correlated equilibrium.
- Strategies, hash power, round duration, rewards, costs, latency, and the learning parameter are all configurable.
- The base assumptions of the model are as follows:
- Using the mining simulator, the researchers examine what potential deviant mining behavior may result from the transaction-fee only model.
- The authors also use theoretical analysis when necessary for certain outcomes.
Results
- If forks exist then Petty Compliance results in higher revenue for miners than Default Compliance and will always be the optimal strategy.
- If Petty Compliance exists within the chain, other miners may pursue higher profits by intentionally forking the chain.
- A “Lazy Fork” may occur when miners fork a chain if the head block is more valuable than the unclaimed transaction fees it leaves behind. This results in unclaimed transactions and may lead to double-spending.
- If miners strategize, increasingly aggressive undercutting behavior evolves, meaning miners may try to extend upon blocks previously completed instead of the current block.
- “An equilibrium exists where all miners use the same undercutting strategy. It induces a growing backlog of transactions” (pg. 8) which may eventually occur when just 0.01% of miners use the equilibrium strategy.
- A minimum of 33% of miners need to undercut for the expected profit of doing so to be greater than that of Default Compliance.
- Selfish mining strategies are more profitable when transaction fees are the sole reward for miners.
- If a miner completes a selfish block, and starts to extend a private chain, the simulation indicates that the greater the selfish-miner’s lead in finding blocks, the disproportionately larger their found blocks will become, compared with if they had no (or a very small) lead.
- Selfish mining can be further “improved” by hiding “small” blocks and immediately publishing “large” blocks.
- Hiding “small” blocks, for example a new block immediately after the previous block was published and before any transactions are announced, with little to zero reward allows for miners to attempt to selfishly mine at (near) zero cost while having a positive probability of building a block lead of two or more.
Discussion
- Under a block rewards design, selfish mining is not immediately profitable. However, under a transaction fee design selfish mining is immediately profitable.
- The existence of Petty Compliant miners in the transaction fee design indirectly improves this selfish mining phenomenon.
- The undercutting attacks discussed undermine the assumption users will be able to increase the speed of transactions by paying a sufficiently large transaction fee.
- “That is because undercutting miners are not looking to maximize the transaction fee that they can claim, and don’t have a strong reason to prioritize a transaction with a high fee.” (pg. 14)
- The researchers limited simulations to either miners who are undercutting or selfish mining, but not both. The potential outcomes of miners conducting both of these strategies may have unexpected results given transaction fee design.
- The researchers see block rewards as “integral to the stability of the mining game” and state “monetary inflation resulting from making the block reward permanent… may be a small price to pay to ensure the stability of a cryptocurrency.”
Applicability
While this is an older paper, the general thesis discussed is important for considering implementing different reward mechanisms. Different incentive structures may be able to strengthen future chain security. Other monetary designs such as perpetual block rewards, EIP-1559 (view related research summary here), metastable mechanism, amongst others, may be such structures. Numerous projects are implementing such mechanisms in an attempt to solidify long-term security within their networks while maintaining incentives for participation.