Strange blocks, "Port Scan" on INVALID state packets from legitimate users
Posted: 16 Jan 2009, 19:33
Guys,
I have a peculiar issue that's popped up since installing and configuring csf/lfd.
I do believe I've followed the instructions and config file directives properly, but maybe I've overlooked something obvious to you.
A little background.
CentOS 5.2-64. Opteron 1216. Physical host (not VPS or virtual guest OS). Generic csf/lfd install, I don't use any sort of control panel software on my servers.
My server has two interfaces, eth[01].
eth0 connects to a private backend network, which is protected elsewhere and needs no filters on my server. This network is lets say 10.10.10.0/30 and this interface is lets say 10.10.10.2. I do not have bogons enabled, and I understand how to enable bogons and still have access on this network. Not a concern for me at this time. Bogons are trapped upstream from me anyhow.
eth1 connects to the public network, which is lets say a 123.123.123.128/29, and it's IP would be lets say 123.123.123.130.
eth1 also has four secondaries statically routed to the 123.123.123.130 address, which exist on secondary interfaces eth1:[0-3]. Lets call this network 123.123.124.240/30.
All my apache VirtualHosts use 123.123.124.241 for incoming traffic. because this is a static routed subnet, all outbound traffic originates from the 123.123.123.130 primary source address. Not a problem.
In my previous firewall configs, I did nothing with INVALID state packets. I was interested to see how csf would deal with this, and to see how many invalid states were coming in.
Now here's my real issue...
Some (not all) legitimate users who hit my largest website are triggering the INVALID rules. Almost all of the invalid trips are happening with DPT=80. lfd decides these individuals are port scanning me and will block them for an hour, after which the user discovers "the site is back up" and starts trying to use it again, and lfd blocks again, ad nauseum, until they get a permanent ban.
Posting the rest of this in a sec, something (not the URLs) in the rest of my post is triggering the forum's new user URL prevention regex... I think I only need 1-2 more posts to get where I can post useful data about the issue.
I have a peculiar issue that's popped up since installing and configuring csf/lfd.
I do believe I've followed the instructions and config file directives properly, but maybe I've overlooked something obvious to you.
A little background.
CentOS 5.2-64. Opteron 1216. Physical host (not VPS or virtual guest OS). Generic csf/lfd install, I don't use any sort of control panel software on my servers.
My server has two interfaces, eth[01].
eth0 connects to a private backend network, which is protected elsewhere and needs no filters on my server. This network is lets say 10.10.10.0/30 and this interface is lets say 10.10.10.2. I do not have bogons enabled, and I understand how to enable bogons and still have access on this network. Not a concern for me at this time. Bogons are trapped upstream from me anyhow.
eth1 connects to the public network, which is lets say a 123.123.123.128/29, and it's IP would be lets say 123.123.123.130.
eth1 also has four secondaries statically routed to the 123.123.123.130 address, which exist on secondary interfaces eth1:[0-3]. Lets call this network 123.123.124.240/30.
All my apache VirtualHosts use 123.123.124.241 for incoming traffic. because this is a static routed subnet, all outbound traffic originates from the 123.123.123.130 primary source address. Not a problem.
In my previous firewall configs, I did nothing with INVALID state packets. I was interested to see how csf would deal with this, and to see how many invalid states were coming in.
Now here's my real issue...
Some (not all) legitimate users who hit my largest website are triggering the INVALID rules. Almost all of the invalid trips are happening with DPT=80. lfd decides these individuals are port scanning me and will block them for an hour, after which the user discovers "the site is back up" and starts trying to use it again, and lfd blocks again, ad nauseum, until they get a permanent ban.
Posting the rest of this in a sec, something (not the URLs) in the rest of my post is triggering the forum's new user URL prevention regex... I think I only need 1-2 more posts to get where I can post useful data about the issue.