Reputation based anti-spam

Now by looking at the Name, email, Website, and IP in a historical fashion. We get to know if this "identity" and deal with it as such. Or more specifically, by looking at the number of valid comments with those parameters, we can add a probability score of a specific comment being spam or not.

Of course the greater the number of valid comments the better the overall results for that specific identity. However, this will need a user to have previously commented. But, I guess that goes without saying since we started this whole article with the word, "reputation". In other words, history.

The question that arises here is, what if the spammers fake specific previously known identities? wouldn't that side track this whole mechanism, even worse, use it to get better spamscore than before using it?

Very good and valid question. In fact, this is one side that I considered. The elements I considered to create a distinguished Identity is Email. Email, is never communicated or disclosed. Although, it's not a big secret, but it's STILL too much work for a spammer to do (figure out the name, email AND website) a specific user is using to comment. But it's possible and therefore EVEN with well known identities other measures will provide additional spam values that will outweigh such an attack

So surface of attack is: New users with little or no history. Attacker who knows the combination of Name, Email and Website of users with good scores (like say, Hani Obaid on this blog).

The hard part to figure out, is how to translate spam-score to an actionable item

Thoughts? issues? concerns? Criticism? let me know what you think. Oh and ask me about your scores if you're interested :)