With all the spoofers appearing lately, how about a feature which sends random strings to servers every now and again then checks the reply. If it says it has the file (impossible) then "ban" that IP from the network.

eg. I send a query for "doifjuerhuifgg" to a server. If it replies "!!!doifjuerhuifgg.mp3" then we know it's a spammer. We can safely send every known virus/trojan to that address.