Firewall's best practice holy cows
- The firewall is security. It is all about the perimeter.
- Take your time to approve a firewall rule set, as this is risk mitigation. There is no need for a set of standard changes, as each change needs to be viewed separately. *
- Don't document. This is insecure as someone will steal it. *
- Don't use virtual firewalls. They are unreliable and more vulnerable.
- Don't use VLANs. They are insecure and leak. *
- Always use two firewalls from different vendors in a cascaded installation. Nothing will ever go wrong twice. *
- Don't allow any UDP or ICMP (even for something as useful as network management.) Out of site is out of mind.
- Don't have geographical fail over's connected at layer 2. Layer 3 is always more secure.
- Don't allow dynamic routing (or for that matter any network related function.) A firewall is not a router. *
- The only way to secure Internet Browsers is to use a forward proxy. Nothing else can scrub traffic.
I know of no major incident that has occurred where one of these holy cows would have been a suitable countermeasure. In reality, these have often been part of the causes of major incidents (which is defined as an incident of severe negative business consequence) or needlessly extended resolution times in the expanded incident life cycle.
Are vendors who provide equipment that don't follow these "best practices", delivering "bad practices?" If we look at virtual firewalls it is my experience that it is easier to administer 10 firewalls with 20 rules each, than to manage 1 firewall with 200 rules. Virtual firewalls make reviewing rule sets easier, and thus by definition are a better risk mitigation.