Angle Explains: Ethereum censorship and mev-boost
Recently, some block producers have started to voluntarily exclude specific transactions from their blocks for other reasons than their validity. This represents censorship at the block level, and could become a serious issue if Ethereum wants to maintain its figure of a neutral and open decentralized network.
Differentiating between several types of censorship
Transactions on Ethereum can be censored at many levels. This censorship can be split into two categories: specific application or contract censorship, and network censorship. Though preventing access to frontend applications or contracts is not desirable from a censorship-resistance point-of-view, they are usually maintained by registered companies that have to comply with regulations.
Ethereum is currently facing some censorship directly at the network level. Network level censorship can be split into three different layers, from least to most critical:
- block producers
- block proposers (validator)
- consensus rules.
If censorship on Ethereum ever reaches consensus rules, it would lose its status of an open and neutral decentralized platform. However, this is not on the table yet.
Today, many Ethereum validators run block producing software that is censoring transactions involving addresses on the OFAC’s list*. This isn’t a validator level issue yet either, as most of them seem to use these block producers because of their better profitability rather than the censorship they apply.
*(US Office of Foreign Assets Control)
The share of block rejecting transactions linked to an address on the OFAC list can be consulted on mevwatch.info as seen below.
State of Ethereum’s censorship today
Flashbots is the company that developed mev-boost, a relay network between block producers and block proposers (ex-miners, now validators). The current debate going on around censorship on Ethereum actually started when they decided to open-source the software and update their client.
It changed two things: 1) it allowed everyone to build their version of mev-boost, actually improving the decentralization of block producing, but 2) they added a censoring of the transactions mentioned above in their client of mev-boost.
It wasn’t the case under Proof-of-Work as block proposers (miners) could include additional transactions in their blocks along with the Flashbots bundle. However, because of the switch to Proof-of-Stake and the new consensus client diversity, mev-boost was updated in a way that builders don’t send bundles to proposers anymore, but block headers. (hear Hasu’s explanation) The big difference here is that bundles could be completed with other transactions to form a block, while block headers can’t.
Flashbots also has more advanced knowledge and skills in MEV, making their mev-boost client the most profitable and favorite one of validators. As they build censoring blocks that can’t be modified by validators, the majority of validators (>60% at the end of October 2022) is effectively publishing blocks censoring transactions involving addresses on the OFAC’s list.
We mentioned that mev-boost has been open-sourced by Flashbots, so that anyone can modify it to not censor any transactions. At the time of writing, only 4% of validators running mev-boost run a non-censoring version of the client. This is likely due to the fact that Flashots’ mev-boost client is still clearly more profitable for validators than the others.
Perspective and possible solutions
Though this touches a fundamental principle of Ethereum, it is to be put in perspective, as Elias Simos from Rated explains in this thread.
First, unless all validators run censoring block-building clients, censored transactions will be included in a block eventually. With the current share of censored blocks, a censored transaction has a > 99% chance of getting included in less than 25 blocks (< 5 min). Though censoring is being performed by some validators, no transaction has been censored at the network layer yet.
Second, the number of OFAC list transactions represented only 0.0025% of all post-merge transactions after a month.
Even though these figures are still pretty low, they could quickly move higher. However, there are already multiple possible ways out of this censorship problem:
- Encouraging more validators to run other relayers not censoring transactions.
- Updating mev-boost to submit bundles instead of block headers, as was the case under Proof-of-work.
- Forcing validators to stop censoring through social slashing.
When looking at the actual figures and the potential solutions, full censorship on Ethereum seems unlikely.
If this situation progresses further and validators start deliberately censoring transactions, then Ethereum users and stakeholders will have to take stand and react if they want to maintain the neutrality of the platform. Community as a whole has the underestimated power to slash targeted validators. It is detailed by Eric Wall in his 30 minutes DevConVI presentation about Social Slashing, where he also tells the story of when a similar situation happened in Bitcoin in 2017.
