Story-time: back in the early days, I knew this guy who ran an IRC server at the ISP where he worked. There was some kid who kept trying to flood him with ping requests (that used to be an effective attack when most people were on a 28.8K connection, and the attacker would be at a university location).
That administrator dove into his IRC client's source code (he was using ircII), and modified the PING response code to return a 64 KB packet instead of a 64 bytes one).
That poor kid quickly went offline after he sent those couple of ping requests at him, and got flooded with 64 KB replies, being sent through that ISP's backbone... That took care of the problem
Back to OP issue: if the "attack" is always the same (a simple connection attempt to a specific port), then just ignore it. The firewall will stop him there. If however he's attempting multiple attack vector on a regular basis, I guess you could block him either through a manual iptables entry in the INPUT chain, or by using the router's Network Service Filtering.