Intrusion Detection Systems mailing list archives
Re: RE: Network Utilization discussion...
From: gdead () shmoo com (Bruce Potter)
Date: Thu, 09 Dec 1999 14:17:50 -0900
tried to send this the other but it didn't make it due to a subscription fubar... Wouldn't it be possible to use one of the layer 4 switches (arrowpoint, etc) to to create a distributed pool of IDS sensors watching a high bandwidth stream? Basically, take the port that you'd be monitoring out of the switch/router/whatever into the L4 switch. Then take and break out the traffic into reasonable parts. With some intelligent segmenting of traffic you could put very specific IDS's against certain streams while still handling large amounts of traffic. ie, if your traffic looked like: Protocol Total Flows Packets Bytes Packets Active(Sec) Idle(Sec) -------- Flows /Sec /Flow /Pkt /Sec /Flow /Flow TCP-Telnet 1159 0.0 29 62 0.0 15.7 14.3 TCP-FTP 6732 0.0 15 69 0.0 5.5 8.3 TCP-FTPD 4864 0.0 17 264 0.0 3.2 6.9 TCP-WWW 198377 0.1 17 1052 2.2 9.3 8.1 TCP-SMTP 17353 0.0 14 280 0.1 4.6 7.0 TCP-X 1 0.0 1 44 0.0 0.0 15.4 TCP-BGP 2 0.0 2 41 0.0 0.7 10.9 TCP-NNTP 1986 0.0 1483 479 1.9 81.1 14.7 TCP-Frag 1 0.0 1 223 0.0 0.0 15.9 TCP-other 56864 0.0 8 498 0.3 4.2 11.6 UDP-DNS 15549 0.0 1 60 0.0 0.8 15.4 UDP-NTP 18011 0.0 1 76 0.0 0.0 15.4 UDP-Frag 1 0.0 1 89 0.0 0.0 15.4 UDP-other 100433 0.0 1 220 0.1 0.7 15.4 ICMP 45478 0.0 19 83 0.5 20.6 15.3 Total: 466811 0.3 18 646 5.4 7.3 11.3 granted this is a low traffic router, but the ratio still works. one IDS box could get all port 80 traffic and the other box would get the rest.. this would scale up to 80Mb/s-ish given the current status of IDS capablity. This setup would require a little intelligence on the backend looking for scans that would span the two interception domains (ie: some attcker going after a php and portmapper attack.. the combined information carries much more weight than one of the events singly). just a thought. bruce Daily security news at http://www.shmoo.com Ron Gula wrote:
What's very clear (at a minimum) from this thread and others like it is: There is no Gigabit or FDDI IDS solution. ISPs who are sporting OC -12s and OC-48s cannot expect Intrusion Detection Systems to work accurately for them, especially if most of the IDS world cannot reliably capture DS-3 utilitization levels.- FDDI is mostly an interface problem. Dragon has been deployed on several FDDI networks through the use of dedicated media converters. Any other packet IDS should be able to do this. Many other NIDS can read directly from FDDI networks. (Read Bob Graham's IDS FAQ) FDDI is also 100 Mb/s and should be able to be monitored by a wide variety of NIDS unless the data rates are in access of 50-60 Mb/s. Once the data goes above those rates it really depends on which NIDS you test, what your data is and how you configure the NIDS. - OC-3, OC-12 and OC-48 interfaces require a "bump on the wire" or a passive tap. That is, the OC link gets plugged into a box that does "passive" IDS before moving the packets or some sort of silvered mirror tap (like a Shomiti tap) is used to pull off the light signal. Of course more sophisticated software is required to rebuild the ATM traffic. The "bump" approach tends to slow down any network traffic and is usually very cost prohibitive. It also usually fought by any WAN engineer because it is a single point of failure. Some high-end products use the passive tap approach and we will be incorporating this into future Dragon Appliance offerings. - We're finding that most of our high bandwidth customers are using full duplex Ethernet (~200 Mb/s), Gigabit Ethernet or IP over SONET. Ron Gula Network Security Wizards
Current thread:
- Re: RE: Network Utilization discussion... Ron Gula (Dec 07)
- <Possible follow-ups>
- Re: RE: Network Utilization discussion... Bruce Potter (Dec 09)
