If there's a public mining pool it's easy to estimate it's computing power. Just take part in the mining pool and compare your own computing power with the percentage of the 50 BTC that you get for each block. Or, alternatively, you can track the BTC that you get from the mining pool back to the block where they were generated, and so you know that this block must have been generated by the mining pool (again, allowing you to use the amount of blocks generated by the mining pool to estimate it's overall computing power).

However, if the attacker does not use a public mining pool, but private infrastructure, things get more complicated: The only way how Bitcoin clients can agree on a common view of the block chain is to treat the highest block number as the authoritative one. A malicious miner can now just generate new blocks and inject them over Tor to some of the Bitcoin nodes. As Bitcoin nodes distribute the blockchain amongst each others, those "malicious" blocks will quickly spread throughout the Bitcoin network.

The only way how you could prevent this problem is some kind of whitelisting approach, where miners need to be manually approved before their blocks are accepted. However, this would totally change the current Bitcoin infrastructure and turn a completely decentralized network into a more centralized one.

This is very interesting. Thank you for the info.

I think a solution to this attack would be a web of trust, like PGP/GnuPG works for verifying signatures.