Fighting DCSpam

yilard · Post by **yilard** » 2003-02-05 12:42

A substring based filtering approach will work for a while until they change the message. There is no problem for them to generate different message over and over until you will be filtering out also innocuous chat. You can win the battle this way, but the war is lost.

I would like to suggest interactive approach involving harrassed users and responsible ops which can help to ban spamming users:

I don't know the dc protocol very well (nearly I don't know it at all) I know that there are script that allow ops to snoop (spy) on user chat including private chat so I think that the PM communication goes through hub.

It should be possible to make a script that will archive all PM communication for a short time (let's say 1-2 minutes). Then if user receives spam it will issue some command like .spam <spamming-nick> and this will trigger another script that will forward PM communcation (those last 1-2 minutes) between user and the spammer to the ops. And its upon ops' consideration whether to ban the user.

I suggested this way to avoid abuse from the side of users to get banned everyone they want.

Hope this is not an overkill

Any comments?

yilard · Post by **yilard** » 2003-02-05 14:47

sorry for posting here, when I look at the post now, I think this should go into Hubs and Scripts section. Any Kind Moderator who could move it there?

ButterflySoul · Post by **ButterflySoul** » 2003-02-06 06:32

I'm not sure what you were refering to exactly... Is it a software/client designed especially to spam ?

Anyway, the clsDCUser class has a property called iChatCharactersInLastMinute which is what I use myself to disconect (not plain filter) any client getting it over 500. Takes care of the chat....

With the Len function, it's pretty easy to keep track of the characters sent via PMs without further analysing the content. It's a bit heavier than relying on the built-in property for chat, but if you use a dictionary, you only need to index the usernames, and assign them a value (i.e. the characters sent during the last minute). Since dictionaries are indexed, you can directly access/update a user's status without scanning through the whole archieved data, as you would have to do if you were to store all PMs in some variable or textfile, then extract a specific conversation from the raw data (which would be quite straining on the hub). I guess you wouldn't want OPs to get all the private messages that happened during the last 2 minutes, because ... err.. well, some private conversations should precisely stay as private as possible =p

ButterflySoul · Post by **ButterflySoul** » 2003-02-06 18:33

Well, if you had a method to disconect them as they enter the hub, it would indeed be better not to post how it works exactly, or the next version of the progie would have a parade against it. That said, there might be a way to distribute the resulting script without ever showing the source, in case something can be worked out with the Microsoft Script Encoder

The method I suggested above doesn't need to be a big secret since it relies on how many character were sent... The only way around it is to not send anything, hence not spam. Of course, it's way less elegant than your approach since the hub (and the users) have to suffer through a few PMs before catching someone.

Could you PM me an addie to the program ? Couldn't find anything on search engines, but I'd like to give it a try on a test hub and see if something can be done -or give a hand if something is already being done-.

A · Post by A » 2003-02-07 13:18

I really don't know much about dc or dcspam don't know if
it's possible what i'm going to write but then again maby you
do.

Isn't there some way to delay it. If the users comes into the
hub that the must be first 2min in the hub before they may
send a pm. and a that the can't send another users immediatly
a pm but that the must wait again a minute with that.

this way it's should take a long while before someone can
send everbody a pm in the hub.

sorry for the bad english, i hope that i made my point

Gadget · Post by **Gadget** » 2003-02-07 16:56

If the users comes into the hub that the must be first 2min in the hub before they may send a pm.

Here is such script.

A · Post by A » 2003-02-08 04:51

Tx for the reply Gadget

Could this be the answer against spam
or does the spam program works different.

!set logintimeout <value> Sets timeout after login before users can chat (currently 2 minutes

and does this feature needs the mod nmdc or can i use this feature
seppartly in a normal nmdc hub.

And i really like to if something like this is available for the Ptokax hub

Gadget · Post by **Gadget** » 2003-02-08 05:48

That should limit the spam so that spammer cant send pm's more than n users, even if he waits that login timeout, which he currently does not. That script should also stop other advertizers and prevent main chat flood.

Script really needs a modified hub with main chat and pm's disabled, script handles those. There is no option to restrict chat with a normal hub software. All you have to change is 4 bytes in binary.

AFAIK, there will be flood/spam filters in Ptokax too.

Sapporo · Post by **Sapporo** » 2003-02-09 23:18

Why not just model it after what the IRC networks use?

For spamming to multiple users, use a "Target Change to fast" type system, just like IRC. If you try to send messages to too many targets in a short span of time, you get a "Target Change to fast (wait 70secs)" message from the server. You can still send messages to windows you currently have open to users and channels but you can't open anymore until that time is up. The server just ignores those requests. If you continue to try and message more users it just keeps adding wait time and continues to ignore you.