It sounds like you need to read some RFCs pertaining to the SMTP protocol. http://www.ietf.org/rfc/rfc2821.txt
During an SMTP connection there is the greeting, acknowledgement, sender, accept/reject, receiver, accept/reject, etc. If postfix is following the rules, the correct action for an address/user that does not exist is fatal/permanent error code 550, user unknown. If you do not need to receive email to your domain 24x7, you can turn of the smtp daemon (postfix), and let all connections to port 25 get refused. I am more familiar with sendmail, but other than cool points of having your own domain/mail server, there are many ways to work around the lack of having one too. Fetchmail, or ETRN for example.
Sorry, I went down a rabbit trail. To answer your question, you need to intercept the part of the SMTP handshake during the RCPT TO portion of the communication, but if you do not respond with a rejection, the other end will time out and try again later (for up to 4 days or more).
Nobody likes being hammered on by hackers. Perhaps a tcpwrapper, a firewall rule, or an RBL is a better solution than what you are actually asking for. You could also write a log parser to watch your postfix log for these failure messages to notify you too, but I would suggest not running it more than once a day for your own sanity's sake.