With greater than 50% of capacity, you can do a lot of bad things.
For ASIC blockchains like bitcoin, a few companies in china dominate the network so they could easily collude and do just this.
For non-ASIC blockchains, you can do it through the cloud. Here is some math i did on it the other day:
digiconomist [0] estimates that current etherum mining cost is 1.3 billion a year, or 3.6 million a day, or 151,000 an hour, or 2,500 a minute.
Multiply by 5 for cloud on demand premiums and you could dominate the etherum network for an entire day for 18 million. You could also do it for free if you can manage to do it with stolen credit cards.
The reality is that even for Bitcoin, the attack, as described, is infeasible. It is possible for miners to collude to censor transactions, but not to create fake transactions, or to somehow steal money from a person and deliver it to others. An attack on that scale would require rewriting the block validity criteria, and no existing nodes or other miners would accept the new criteria, which would result in the chain forking, and the main chain simply losing a lot of hash power, which would be working on the fork that contained the change the validity criteria.
For ASIC blockchains like bitcoin, a few companies in china dominate the network so they could easily collude and do just this.
For non-ASIC blockchains, you can do it through the cloud. Here is some math i did on it the other day:
digiconomist [0] estimates that current etherum mining cost is 1.3 billion a year, or 3.6 million a day, or 151,000 an hour, or 2,500 a minute.
Multiply by 5 for cloud on demand premiums and you could dominate the etherum network for an entire day for 18 million. You could also do it for free if you can manage to do it with stolen credit cards.
https://digiconomist.net/ethereum-energy-consumption