How do you
implement IDS (network based) in a heavily switched environment?
The difficulty of implementing IDS into a switched environment stems from the basic
differences between standard hubs and switches. Hubs have no concept of a connection and
thus will echo every packet to every port on the hub, excluding only the port the packet
came in on. A switch however is based on connections, when a packet comes in a temporary
connection in the switch is made to the destination port, and the packets are forwarded
on. So in a hub environment we can place our sensors almost anywhere, while with switches
specific workarounds must be used to assure the sensor is able to see the traffic
required.
The current options for this are TAPS, Hubs and Spanning ports. A spanning port configures
the switch to behave like a hub for a specific port. For instance in Figure-1 we wish to
monitor the connection between the switch and the Resource Machine. To do this we tell the
switch to span the data from the resource machines port to the IDS port. This can be done
with Transmit Data, Receive Data or both. Some current switches cannot be relied on to
pass 100% of the traffic to the spanned port, so attacks could go un-noticed even when the
IDS system is configured to look for the attack. Switches also only allow one port to be
spanned at a time, so monitoring multiple machines can be difficult or impossible.
Using Hubs or TAPS is a very similar solution, the hub or tap is placed
between the connection to be monitored. This is usually between either two switches, a
router and switch, or a server and switch, etc. In Figure 2 a hub has been placed between
the resource machine and the switch. This allows traffic to still flow between the switch
and the Resource while the properties of the hub cause a copy of the traffic to be copied
off to the IDS. This, like the span port is only suitable for single machines. Multiple
machines on the hub would cause network problems and remove the benefits of a switched
solution. In addition, to get a fault tolerant hub would increase the cost of the solution
dramatically. Taps are by design fault tolerant having the main connection (i.e. the
connection between the resource and the switch), hardwired into the device, preventing
failure.
Figure 3 shows a tap monitoring a single Resource machine. The tap is
unidirectional and passes traffic from the switch and resource machine to the IDS only.
This prevents traffic passing from the IDS to the switch or Resource machine nor can
traffic be directed at the IDS. Since the Tap is unidirectional we can route the traffic
from several taps back to a hub to be monitored by the IDS system, without causing network
problems, this is shown in Figure 4.
www.shomiti.com, www.ods.com, are makers of taps
Brian W Laing
Internet Security Systems
< Previous : Next >
Click here to return to the FAQ index |