Script to scan already arrived email for keywords on Linux

Hi, I'm looking for someone who has experience with the inner working of mailservers and (perl) scripting in Linux to write a (perl) script to scan already arrived email for keywords. Background: Our mailserver only processes incoming emails, no email sending is taking place. On an average day about 600K email arrive at the server. This 600K is received on about 40 local catchall mailboxes that are forwarded to 4 central emailaccounts. These 4 central emails accounts are accessed by 4 different pop3 clients at 1 minute intervals. Of all the incoming mail only mail that contains specific keywords inside urls within the body of the mail are of use to us. Currently the 4 central pop accounts receive too many unwanted mails to function proporly and regular spamfilters dont fit our purpose. Project: Thus what is needed is a script that continuously or semi-continuously (1 minute interval for example) checks all mail that has arrived on the server (for example in /var/spool/mail or /home/usrx/mail/new, you determine the best location) for specific keywords (whitelist) inside urls within the body of the mail. If one or more words from a whitelist is present then the mail kept, otherwise the mail is deleted. - The server is an out of the vps with 3 cores (3x AMD Opteron 6174 @ 2.2GHZ) with 4GB of ram with CENTOS, WHM/CPANEL and the usual programs installed. IF you suggest different hardware let me know. - You are free to select & install any mailserver (dovecot, courier, etc) or other software as you see fit. The server is dedicated 100% to this project. - The 40 catchal domains will change from to time, the 4 central emailaccounts will not. Please make your solution supports this. - Speed of the script is essential as mail will be coming every second of the day! The script should check all mail present and incorporate a fast mechanism to refrain from scanning email multiple times. Maybe a temp direcory, renaming of files, this is up to you. No unscanned email should be downloaded by the pop3 client. - Just to let you know. Spamassisin had already been tried but result in very high load. Also ASSP has been tried unsuccesfully (cant search within the body of the email). Realtime scanning at arrival time is too cpu intensive. - I want to receive the sourcecode and some documentation for when I decide to move to another server. - A simple web interface is needed for: 1. monitoring the number of emails currently located on the server (to make sure no email piles up) 2. Entering the words for the whitelist. 3. monitoring the serverload. - Example: if the body of the mail is like this: "Hello user x, please visit this link to start: <a href="[login to view URL]">click here</a>" and the whitelist include the word 'verify' then this mail should be kept. Goodluck with bidding