Guides

How to Check a Firewall Change Request Before Approval

firewall change requestchange intakerulebase checkticket automation
Engineer checking a printed list of firewall rules line by line with a pen, a rack-mounted firewall with status lights behind him

Check a firewall change request before approval in two steps: confirm the request is complete, then test it against the live rulebase. A complete request answers nine questions, from source, destination and service to business reason, owner and an end or review date. The rulebase test returns one of four results: already allowed, blocked by an earlier rule, conflicting with another rule, or too broad. A request that triggers none of the four can go to the approvers. Incomplete, already allowed and too broad requests go back to the requester first. Blocked and conflicting results stay with the request as findings for the engineer and the approver.

FwChange checks every request before it reaches an approver: it reads the request from a ticket system, checks every field and tests it against the firewall's current rules. Book a demo.

The reason to do it before approval is cost. An approver who reads a request with "the app servers" as the source and no reason either guesses or sends it back later. NIST SP 800-53 control CM-3 asks organizations to "Review proposed configuration-controlled changes to the system and approve or disapprove such changes", and NIST SP 800-41 Rev. 1 says that "Firewall rulesets and policies should be managed by a formal change management control process because of their potential to impact security and business operations". A review is only as good as what reaches it. This post is about the gate in front of the review.

It is the stage between the ticket and the approval. The engineer's analysis of what a change does to traffic is in firewall rule impact analysis. Who may sign is in approval chains and separation of duties. Adding network context to the reviewer's screen is in NetBox context for firewall rule reviews.

What must a complete firewall change request contain?

A complete request contains nine things: source, destination, service, action, the firewall or zone it applies to, the business reason, a named owner, an end or review date, and a priority with the wanted window. Free text for any of the first five invites "the app servers" and "the usual ports". Use separate fields and reject a request when one is empty or vague.

FieldComplete meansSend back when
SourceNamed hosts, a subnet or an existing object"The app servers", "internal", or any with no explanation
DestinationNamed hosts, a subnet or an existing objectA site or team name instead of addresses
ServiceProtocol and port or port range"Web", "the usual ports", or a range where one port is needed
ActionAllow or deny, and the directionMissing, or contradicts the reason
Firewall or zoneThe device or zone pair the rule belongs onNot named, or not in your inventory
Business reasonThe application or project, and the requirement behind itEmpty, or only "needed for access"
OwnerOne named person who answers for the rule laterA team mailbox, or a person who has left
End or review dateA date inside the maximum your policy allowsBlank, in the past, or "permanent" with no review
Priority and windowThe urgency and the wanted change windowMissing, so the approval order cannot be set

Reason and owner are not paperwork. NIST SP 800-41 Rev. 1 says that "Filling in such a comment is important for others to determine why a rule was made." A reason and an owner captured at the request are the comment on the rule. The German BSI module NET.3.2 Firewall points the same way in requirement A2: "Die getroffenen Entscheidungen sowie die relevanten Informationen und Entscheidungsgründe MÜSSEN dokumentiert werden" (our translation: the decisions taken and the relevant information and reasons must be documented). The module does not list request fields, so the nine above are our recommendation.

The end or review date gives each rule a reason to be looked at again. NIST SP 800-41 says a policy review "can uncover rules that are no longer needed". A date turns that from a periodic sweep into a per-rule trigger.

How do you check the fields before touching the firewall?

Check the fields mechanically first, because a rulebase test on a malformed request returns noise. The list below is guidance for building your own check, not a description of what FwChange does. Each check is a yes or no that a script can run, and each failure produces one specific message for the requester. Only a request that passes every field check goes on to the rulebase test.

  • Every required field is filled.
  • Addresses parse as a host, a range or a CIDR block, and named objects exist.
  • Ports are numbers between 1 and 65535 and the protocol is named.
  • Any in source, destination or service is flagged and needs a written reason.
  • The firewall or zone exists in your inventory.
  • The owner exists in your directory and is active.
  • The end date is in the future and inside your policy maximum.
  • Source and destination are not the same, and the action matches the reason.

A script cannot judge the reason. It can only see that one exists. A human reads the reason at approval, and the check's job is to make sure there is something to read.

What are the four tests against the live rulebase?

Test the request against the rules the firewall runs today, not against a spreadsheet or last quarter's export. For the source, destination and service in the request, ask which existing rule matches, then sort the answer into one of four results. Expand groups to addresses before you compare, because a rule that names a group and a request that names a host will not match as text.

ResultWhat it meansWhat happens next
Already allowedAn existing rule already permits this flowNo change needed. Close the request and name the rule. If the user still cannot connect, the fault is elsewhere: routing, DNS, the host
Blocked by an earlier ruleA rule above the place for the new rule matches first, so the new one would never be reachedThe request is valid but the position or the earlier rule needs a decision. Flag it for the approver and the engineer
ConflictingAnother rule covers part of the same traffic with the opposite action or different settingsThe new rule would change or contradict it. The approver must see both rules side by side
Too broadThe request uses any, a whole subnet or a port rangeReturn to the requester: which hosts and ports does the application need?

The order of rules is the reason for the second result. NIST SP 800-41 Rev. 1 notes that "many firewalls check traffic against rules in a sequential manner until a match is found". A request that is correct on paper can be dead on arrival if an older rule sits above it.

"Too broad" needs a definition before you automate it, and a mechanical one works best. Our working definition: a request is too broad when it uses any as source, destination or service, covers a whole subnet where hosts could be named, or lists a port range instead of the specific ports. The check flags these patterns, and a person decides whether the width is justified.

The four tests look at the rulebase as it is now. They do not replace the engineer's analysis of knock-on effects on other traffic, which is covered in firewall rule impact analysis. Some firewall vendors now ship their own check before a change is made, for example Palo Alto Networks with the Policy Analyzer in Strata Cloud Manager. Whether a change board should accept a platform's grade on its own change is the subject of our post on vendor AI and impact analysis.

What do you send back to the requester?

Send one message per request that names every failed item, says what is needed, and leaves the ticket in a status that shows it was returned. A message like "incomplete" costs a round trip. A message that quotes the field and says what a good answer looks like often closes in one reply. Keep the wording in the requester's vocabulary, since many requesters are not firewall people.

FindingWhat to write
Vague source"Source says 'app servers'. Please give the host names or the subnet."
No reason"Please name the application and the project or requirement this access is for."
No owner or expiry"Who answers for this rule? Until when is it needed?"
Already allowed"Rule [name] already permits this traffic. Please check routing, DNS and the host, then reopen if it still fails."
Too broad"You asked for [wide field]. Which hosts and ports does the application actually use?"

Two of the four results, blocked by an earlier rule and conflicting, are not the requester's fault. Do not send those back. Attach them to the request as findings for the approver and the engineer. This split is our recommendation, and it keeps the requester from being asked questions only a firewall person can answer.

How can you automate the check from the ticket system?

If you build the check yourself, automate it as a gate with four parts: a form with structured fields, a trigger when the ticket is submitted, a service that runs the field checks and then the four rulebase tests with read-only access, and a write-back that sets the ticket status and posts the findings. The check gates the approval and never approves anything itself.

  • Form. Separate fields for every item in the table above, with dropdowns where the set is known (firewall, zone, owner). Free text only for the reason.
  • Trigger. A webhook on submission, or a poll every few minutes if webhooks are not available.
  • Check service. Runs the field checks first. If they pass, it queries the rulebase through a read-only account, or against a fresh export, and records the result with the time and the rulebase version it read.
  • Write-back. One comment with the findings, and a status of "returned" or "ready for approval".
  • Expiry of the result. Rules move. Re-run the test when the ticket reaches approval if the rulebase has changed since the check.

Keep the credentials read-only. A check that can edit the rulebase is a change tool, and it would need its own approval chain. Keep the check result in the ticket history, not only in a log file, so the auditor sees it next to the approval. A change made on the firewall with no request at all is a different problem, covered in drift detection for unauthorized firewall changes. The whole process around this gate is in the firewall change management guide.

How does FwChange run this check?

FwChange reads the request from a ticket system and checks every field. An incomplete or wrong request goes back to the requester. A complete one is tested against the firewall's current rules, with the four results above: already allowed, blocked by an earlier rule, conflicting, or too broad. Then it is routed to the approvers by priority, and nobody approves their own change.

  • FwChange keeps an audit trail that you can export for an auditor.
  • FwChange uses read-only firewall accounts and never changes a firewall. Your engineers make the change.
  • It is self-hosted on one virtual machine in your own network and needs no internet access for the core workflow.

Book a demo.

Frequently asked questions

Does FwChange change anything on the firewall?

No. It uses read-only accounts and never changes a firewall. Your engineers make the change.

What does FwChange check on a request?

It reads the request from a ticket system and checks every field. It sends an incomplete or wrong request back to the requester. It tests a complete request against the firewall's current rules and reports whether it is already allowed, blocked by an earlier rule, conflicting or too broad.

Where does FwChange run, and does it need internet access?

It is self-hosted on one virtual machine in your own network. The core workflow needs no internet access.

Checklist: the request gate

  • The request form has separate fields for source, destination, service, action, firewall or zone, reason, owner, end or review date, and priority.
  • Every field passes a mechanical check before anyone reads the request.
  • Any in a field needs a written reason.
  • The owner is a named, active person.
  • The end or review date sits inside your policy maximum.
  • The request is tested against the live rulebase, with groups expanded.
  • The result is one of four: already allowed, blocked by an earlier rule, conflicting, too broad.
  • Incomplete, already allowed and too broad requests go back with a specific message.
  • Blocked-by-earlier and conflicting results are attached for the approver and the engineer.
  • The check uses a read-only account and approves nothing.
  • The check result and its time are stored with the ticket and re-run if the rulebase changed before approval.

How we checked this

On 22 November 2026 we read NIST SP 800-53 Rev. 5 control CM-3 as reproduced on csf.tools, NIST SP 800-41 Rev. 1 (September 2009) and the BSI IT-Grundschutz module NET.3.2 Firewall (Edition 2023). The passages we read do not list request fields and do not define the four results. The nine fields, the four results, the definition of "too broad", the send-back split and the automation pattern are our own recommendation.

Book a demo.

FW

FwChange

Firewall change management

Methodology and software for firewall change management.