> Compare this to PoW, where only 51% of hashrate needs to be participating/complicit to censor certain transactions, the other 49% have no way of fighting back, and the only tool the community has to fork away is to change the PoW algorithm.
This is just such a small part of the whole equation.
1. They can't continuously attack with this, as you need to expend energy as long as you want to continue with the attack.
2. Even if you have majority hash-rate, you can't change the rules of the system (this has already happened)
1. Under PoW, a 51% attacker can continuously attack profitably. All the energy they expend gets returned as mining rewards just like normal, and they can potentially even (up to) double their rewards if they censor the other 49%. The only way to stop this aside from changing the PoW algorithm is to physically locate and seize the mining rigs.
2. Majority stakers can't change the rules of the system either.
> All the energy they expend gets returned as mining rewards just like normal
Is that because the mined blocks themselves are still valid (i.e. the hashes check out), regardless of the presence of an attacker? But then how does it help the attacker in an economical sense? Doesn't the rest of the network know that the new 51% blocks are tainted, so to speak?
>Is that because the mined blocks themselves are still valid (i.e. the hashes check out), regardless of the presence of an attacker?
Yes. A censorship attack is just that - it's about mining valid blocks that ignore transactions from a certain party, and/or ignoring blocks from other block producers.
> But then how does it help the attacker in an economical sense?
If an attacker has 60% and censors all other block producers (40%), those other block producers won't be able to earn any mining rewards because teh 60% can continuously orphan the blocks they produce, putting them on non-longest-chain (invalid) forks.
In writing my last reply I forgot that difficulty adjustment only happens every 14 days (I can't remember how long it is with Ethereum's PoW, but let's take Bitcoin's PoW as an example), so it's likely that the attacker wouldn't earn any more mining rewards than an honest miner at first.
When that difficulty adjustment happens though, if the miner had been censoring for a good part of the past 14 days, the network would adjust to "consider their 60% the entire hashrate of the network", thus the attacker would start earning the other 40% of rewards for themselves in addition to the rewards they are supposed to earn.
> Doesn't the rest of the network know that the new 51% blocks are tainted, so to speak?
It becomes a very messy, subjective problem. For every block, you would have to prove whether the block producer is part of the censorship attack or just has an incomplete view of the network, which is an intractable computing problem without some sort of subjective heuristic requiring multiple network viewpoints and trusted parties.
This is just such a small part of the whole equation.
1. They can't continuously attack with this, as you need to expend energy as long as you want to continue with the attack.
2. Even if you have majority hash-rate, you can't change the rules of the system (this has already happened)