A reader wrote in today asking about egress filtering.  It seemed like a perfect topic considering it is one that seems to spark emotions depending on your role in your organization.  I've heard arguments from both sides.  Generally the lines are divided with the Network Group who deem it is to hard to do (i.e. lots of acls and/or rules, etc.) and the Security Group who says it needs to be done.  Here are some questions to consider with egress filtering:

1.  Is it necessary to do egress filtering? The short answer is yes.  You can not defend your company's information and network without doing egress filtering.  

2.  Where do you need to do egress filtering on your network?  This, I think, is one of the real questions.  Most people think at the one point of the network that allows access to the internet such as the DMZ.  But what about internal "egress" filtering?  If your network is divided up based on business lines and there are key areas such as finance or code repositories etc. would you want to do egress filtering there?  I would submit that the answer is yes.  The information may not be going to the internet but to an insider threat.  Or a compromised box is collecting the information internally on your network and that box has access to go to the internet.  You would hope that your egress filtering would catch that, but it really depends on the kinds of filtering you're doing. 

3.  What devices can filter for you?  The concept of egress filtering has really broadened over the years and technology has driven that.  Still, many people associate egress filtering with just routers and firewalls.  But many other tools out there  have egress filtering capabilities such as web proxies and email gateways.  These are actually the right tools in many cases to do filtering.  ACLS and firewall rules serve their purpose, but in most cases its limited.  Firewalls, for example, allow you to create rules that control what traffic can go outbound.  But if its a stateful inspection firewall, its inspection capabilities are very limited for application layer inspection.  I'm refering to stateful inspection firewalls that allow for limited deep packet inspection for certain protocols such as HTTP.  A web proxy can actually give you way more granular control over your HTTP traffic.

4.  Isn't there such a thing as filtering to the point of diminishing returns?  Yes, there is.  If you are trying to filter everything and on every device, then you may hit diminishing returns for the amount of work you are creating for yourself.  If your ACLs and firewall rules are so many that you don't know what they do, then you have a problem.  You can actually be allowing traffic to pass when you thought you were blocking it (depending on the device type and firing order of the acls/rules).  For example, if you are depending on all your routers to do all your egress filtering, then you may be creating more work for yourself.  Your router may be able to filter out at a higher level the traffic and allow other devices to be the "real" egress enforcer.   

The bottom line is that there is no "one size fits all" for egress filtering on your network.  You have to look at your network design, know what your network traffic consists of (I can't emphasize that one enough), figure out where you need egress control points on your network and identify what device is best suited for it.  If you realize you don't have what you need, add it to your risk management and mitigation plan.  I also really like building a traffic flow diagram that helps me see (at a high level) who is allowed to talk to who and on what port and protocol.  This also helps me determine where I need control points.  This is only a high level discussion on egress filtering.  If you have something you do or tool you like to use, please let us know and we'll update the diary.