> > I'm currently working on getting policyd rate limiting, and outbound spam > > filtering put together, hopefully by the end of the week, but in the mean > > time > > I'm hoping there is a way to limit sending auth sessions to a time limit to > > help stop the bleeding. > > http://www.postfix.org/TUNING_README.html#conn_limit > http://www.postfix.org/postconf.5.html#smtpd_client_message_rate_limit > http://www.postfix.org/postconf.5.html#smtpd_client_recipient_rate_limit > > Wietse
Thanks Wietse, This is really helpful and exactly what I was looking for. I set it up with a 60s time out and set the message and client limits to a really high rate to helps set a baseline. After a day of business use, the vast majority of users are below 5/6 per min. I have a couple NATed users around 40/min. Are people setting a long anvil_rate_time_unit (like 12 or 24 hours) and white listing users that would normally go over or trying to rate limit with a 60s time_unit? I'm sure it varies per network but I'm thinking that most users for me are well below 1200 messages/24 hours and a compromise that only resulted in 1200-3600 spam messages would be better than what I do now (taking into consideration that most compromised accounts send from a half dozen IPs). I'd love to here other people's thoughts on whether longer anvil_rate_time_units or lower per time_unit messages are a better way to go. Ether way, I've pushed most of the spam messages from the last 2 weeks through amavis and I think they'll all be blocked with out bound filtering, but I think this is a good intermediate step that will also lower overhead on outbound filtering systems. Thanks for your time, -- Pete
