Ethereum's New DeFi Firewall Has a 2-Builder Problem
A security tool that lets DeFi protocols block hostile transactions before they land sounds like a win. It also means someone gets to decide what counts as hostile, and on Ethereum that decision routes through a block builder market with two dominant players. The tool isn't the risk. The concentration is.
Ethereum's newest security tool isn't the scary part. The scary part is that it needs a couple of companies to actually work, and that couple fits in a minivan.
Phylax built a system that lets decentralized protocols screen out transactions they believe are harmful before those transactions settle. On paper that's a firewall for DeFi. In practice it's a veto. And every veto needs someone holding the pen.
That's the whole story. Not the code, not the audit, not the governance forum thread. The pen.
How it works, and why that matters
The pitch is straightforward. A protocol plugs in, defines what a malicious transaction looks like, and gets a chance to stop it before it executes. Given how much money has drained out of DeFi in the last few years, you can see the appeal. The Bybit breach in February 2025 ran to roughly $1.4 billion in ETH, and that wasn't even a smart contract bug, it was an operational failure. Billions more have walked out the door through exploits, oracle manipulations, and key compromises since.
So a pre-execution blocklist isn't a solution looking for a problem. The problem is real and expensive.
Here's where the infrastructure enters the room. Screening a transaction before it settles means someone has to see it first, and on Ethereum that someone is the block builder. Since the Merge, block production has split into a proposer side and a builder side, and the builder side is where the concentration lives. Titan Builder and beaverbuild have at various points cleared more than 80% of Ethereum's blocks between them. Two firms. Most of the blocks.
That's not a conspiracy. It's just the economics of specialized block building, where the fastest and best-capitalized operators win the auction over and over. It's the same story as ASIC manufacturing, or cloud hosting, or any other business with a steep efficiency curve. Mining is an energy business that happens to produce bitcoin. Ethereum's version is a block space business that happens to produce finality.
And you know what else lives down at that layer? Censorship. In the two years after the Tornado Cash sanctions in August 2022, OFAC-compliant relays routinely handled more than half of all blocks. That share eroded through 2024, and after the Treasury lifted the sanctions in March 2025 it collapsed toward zero. The market fixed it. But it fixed it slowly, and only because a legal constraint went away. Nothing about the underlying structure changed.
So when a security product adds a screening layer, the question isn't whether the screening is well-intentioned. It's who operates the filter and how much of the network they touch.
The case for the other side
Let me steelman the bullish read, because it's stronger than the cynics admit.
Block building is permissionless. Anyone can build a block. If a builder starts censoring transactions for reasons its users don't like, proposers can route around it, and builders that drop good fee-paying transactions are leaving money on the table. The auction punishes censorship on its own. That's the design, and it's held up better than most people predicted.
The screening tool is also opt-in. A protocol chooses to use it, the way a bank chooses to run fraud detection. If the criteria are published and the protocol's governance can reverse a block, that's not censorship, that's risk management with an audit trail. Users who hate it can use a fork or a different protocol. DeFi has never lacked for alternatives.
And there's a hard-nosed commercial argument. If your protocol gets drained, you don't have a governance debate, you've an empty contract. Security features attract deposits. Deposits attract liquidity. Liquidity attracts more deposits.
Fair enough. Now the problems.
A blocklist is a target. Publish the criteria and attackers will engineer transactions that slip under the threshold. Keep the criteria private and you've built a black box that decides who can transact. Those are the two doors, and neither one is great.
There's also a legal trap sitting underneath all this. Once a protocol starts actively screening, it's arguably taken on a duty of care. Regulators tend to notice when you build the compliance tool yourself. Ask the relays how that math worked out.
And the concentration problem doesn't go away because the tool is opt-in. A veto that lives at the protocol level is fine. A veto that depends on two builders in overlapping time zones is a single point of failure with a marketing budget.
The verdict
I'll take the tool. I won't take it quietly.
Pre-execution screening is going to become standard, the same way circuit breakers became standard in equities. Users want it, protocols need it, and the alternative is watching another nine figures leave a lending market in a single block. Fighting that trend is a losing position.
But the tool should come with two pieces of paperwork. First, every protocol running a blocklist should publish its criteria and its false positive rate, in public, on a schedule. Opt-in censorship with disclosure is governance. Secret censorship is a backdoor with better branding.
Second, and this is the part the security crowd keeps skipping, watch builder market share the way miners watch difficulty. Follow the hashrate on Bitcoin, sure. On Ethereum the number that matters is how many independent builders are producing blocks, and that number has been too small for too long. Difficulty adjustment is the market nobody watches. Builder share is the one nobody even thinks to check.
The economics here are tighter than people think. Phylax didn't create Ethereum's concentration problem, and it's not responsible for fixing it. But any product that adds a gate had better be honest about how narrow the corridor around that gate really is.
Right now it's narrow enough that two companies could close it. That's the thing to fix. Not the firewall.
Explore More
Key Terms Explained
The first cryptocurrency, created in 2009 by the pseudonymous Satoshi Nakamoto.
A bundle of transactions that gets permanently added to the blockchain.
Following the laws and regulations that apply to financial activities, including crypto.
Not controlled by any single entity, authority, or server.