Firewall Rule Audit Checklist: Find Shadowed Rules First

Use this firewall rule audit checklist to find shadowed, redundant and any-any rules before an auditor does. Practical steps from our field work.

Most firewall policies look tidy on paper and then fall apart the moment you read the device itself. We see this constantly in the field. The policy document says one thing, the running configuration says another, and nobody on the team can explain why half the rules are still there. A firewall rule audit checklist is how you close that gap before an auditor or an incident response team finds it for you. Start with the policy as it is actually enforced, not as someone once intended it to be.

What a Firewall Rule Audit Has to Cover

A proper audit answers a handful of questions about every single rule. Is it still needed? Does it sit in the right position in the order? Does anyone own it? Is it logged? Was the change that created it recorded anywhere? If a rule fails any of those tests, it goes onto a review or removal list until someone can justify it. That sounds strict, but a rule that cannot be justified is usually a rule nobody will defend during an incident.

The core checks we run are shadowed rules, redundant rules, any-any rules, missing owners, expired rules, and missing change records. Each one points to a different kind of risk. A shadowed rule is a logic error. An any-any rule is an exposure. A rule without an owner is a governance failure. Keep them separate in your tracker, because each one gets fixed in a different way.

How Rule Bases Grow Unchecked

Rules almost never get removed on purpose. A project goes live and someone opens a port for a vendor. An outage gets a temporary allow that is never reverted. A migration finishes and nobody goes back to clean up the old entries. Give it a few years and the base holds thousands of lines that no one wants to touch, because touching them might break something.

The trouble is that a rule which looks harmless on its own can quietly shadow a stricter rule further down the list. That interaction is hard to spot by reading line by line. It is also why we keep finding it in customer environments during reviews, usually in places nobody thought to look. Bases of this kind are the hardest to reason about, and they are exactly where an audit pays back the most time.

Export the Policy and Capture a Baseline

Before you change anything, export the running configuration or the policy objects from the device or management platform. Record the software version alongside the export. The same command can return different fields or behave differently across releases, and you will want that context later when someone asks why two reports disagree.

Capture hit counters at the same time. A hit counter shows how often a rule has matched since the counter was last reset. A zero can mean the rule is dead. It can also mean the rule serves a quarterly batch job or a disaster-recovery path that has not fired yet. The baseline lets you separate those cases later, which is the whole point of collecting it now.

On a Cisco ASA style device, a reasonable starting point is show access-list | include hitcnt. Confirm the exact syntax for your OS version first, since output and options shift between releases. Store the export as read-only and label it with the date, the device name, and the person who pulled it.

Finding Rules That Can Never Match

A shadowed rule never gets the chance to match, because an earlier rule already captures all of its traffic. Think of a doorman who checks the first guest list and never reaches the second. A redundant rule is related but different. Its action and scope are fully covered by another rule with the same effect, so removing it changes nothing about how traffic flows.

Walk the rule order from top to bottom. Flag any rule whose source, destination, and service fall entirely inside an earlier rule carrying a conflicting action. Most vendors ship a policy analysis or shadow-check feature, and it saves hours on large bases. Still, confirm the findings by hand on your critical segments. The tool only knows the model it was built around, and a rule tied to a dynamic object or a time window can look very different in practice.

Hunting Down Any-Any and Overly Broad Rules

Flag every rule where the source, destination, or service is set to “any” without a written justification. In our experience these rules fall into three rough groups: leftovers from a troubleshooting session, legacy rules whose purpose has been lost, and deliberate allows for trusted internal paths. Each group needs different handling.

Troubleshooting leftovers are the easiest. Confirm with whoever requested them, then remove them. Legacy rules take more digging. Trace each one back to a system or an owner before you touch it, because deleting a rule nobody understands is how weekend outages start. Intentional trusted paths should stay, but they need a documented reason and a ticket reference attached.

Where you can, replace broad rules with explicit object groups. A group named for the application is far easier to audit than a raw IP range. Log every change against a ticket so the reasoning survives after the person who made it has moved on.

Who Owns a Rule, and When Does It Expire?

Every rule should carry three things: an owner, a business purpose, and a review or expiry date. If your platform supports comment or metadata fields, use them directly on the rule. If it does not, keep a controlled register that maps rule IDs to owners and review dates. Either works, as long as someone actually maintains it.

We usually apply a simple rule of thumb: a rule with no owner after two review cycles goes onto a removal candidate list. That can feel aggressive, but an unowned rule is an unexplained hole in the perimeter, and auditors will ask about it first.

Confirming That Logging Works Where It Counts

Check that deny rules and rules on trust boundaries actually generate logs. Then confirm those logs reach a central store with retention that matches your policy. A logging option that looks enabled in the console can still fail silently under high volume, so do not take the setting at face value.

Send a test packet through a logged rule and query the log store for it. If the entry does not appear, you have found a blind spot. Write the logging standard down, so an auditor can compare it against the live configuration instead of relying on your word for it.

The Change Records an Auditor Will Actually Sample

For each change that comes out of the audit, record the rule ID, the state before and after, the ticket, the approver, and the date. Keep evidence that the change was tested, such as a test plan or the output of a dry run.

Auditors rarely review every change. They sample a handful and trace each one back to its paperwork. A short, traceable record beats a long list of fixes with no documentation behind them.

Manual Review Versus Tooling

Manual review of exported rules is workable for a small policy. Once the base reaches a few thousand entries, it starts missing shadowing across the full order, and that is where tooling earns its place. Vendor analysis tools are faster, but they rely on the vendor’s own model of the rule base, so validate what they report against a sample you check yourself.

We compared both approaches against the same rule set on a recent engagement. The manual pass caught more context-dependent problems, such as rules tied to a retired subnet. The tool caught more pure ordering conflicts in less time. Running both gave the most reliable result, and we now treat them as complementary rather than competing.

Mistakes We Keep Seeing

The most common error is deleting a rule because its hit counter reads zero. Seasonal traffic and disaster-recovery paths often sit idle for months. Check each candidate against the calendar and the recovery runbook before you remove it.

The second is reordering rules during an audit without a rollback plan. Always keep the prior configuration so you can restore it quickly if something breaks. Also, do not trust a shadow report until you have confirmed it supports your exact platform version. Finally, remember that cloud security groups and host-level firewalls often sit outside the audit scope. They need the same checks, or the review leaves a gap that attackers will find.

Frequently Asked Questions

How often should a firewall rule audit run?

Most teams run a full audit at least once a year and a lighter review each quarter. Schedule an extra pass right after a major migration or a security incident, because that is when rules accumulate fastest.

Is a zero hit count enough to delete a rule?

No. Zero hits across your baseline window is a signal to investigate, not proof that the rule is dead. Confirm with the owner and check for seasonal or recovery traffic before you remove anything.

Which rules do auditors look at first?

They usually start with any-any rules, gaps in deny coverage at trust boundaries, and rules with no owner. Those three reveal weak control quickly, so clean them up before anything else.

Can the audit run without downtime?

The export and analysis steps are read-only and need no downtime. Only the removal and reordering steps carry risk, so schedule those in a maintenance window and keep a tested rollback ready.

Do cloud firewall rules need the same checks?

Yes. Security groups and network ACLs have the same shadowing, ownership, and logging problems. Include them in the same register so the review covers the whole traffic path, not only the on-premises firewall. For a broader view of how this fits into segmentation work, see our notes at MONA Media, and browse more of our cybersecurity coverage for related guidance on zero trust controls. If your team is also planning the wider rollout, our write-up on mTLS in hybrid cloud covers how identity-based controls complement rule-based ones.

Related topics