RE: [NeoStats-Devel] FloodServ
"M" <[email protected]>
| Newsgroups | gmane.comp.neostats.devel |
|---|---|
| Message-ID | <[email protected]> |
Justin Hammond wrote: > Some of the improvements that I had thought about was > extending it beyond the channel layer and move it to the IRCd layer. > Eg: > Add support for monitoring a large number of connects from > Class C networks (Hybrid TCM does this, but its usefullness > from the perspective of zombied machines might be limited now) > > Add monitoring for the average number of connects on each > IRCd (eg, if my daily average connects is 1 connect every 10 > seconds, and it all of a sudden jumped to 10 connects, we > might have a inbound flood.) There might not be much you can > do, but at least give the opers some warning... > > Monitoring the channel joins/parts of users. Typically some > of the spam bots or stats bots that are out there will cycle > each public channel very quickly... This could be useful. If we take the webcamspam as an example, it joins channels in order of the raw output of /list so a /join chan1, /part chan1, /join chan2 etc would be sufficient to detect such bots. However, this would mean we would have to maintain a raw list of channels via a client list command unless we can determine an algorithm to identify the first few channels that would be output via raw. I assume it could be determined from TS but have not looked into it. Mark.