Uncia
FR / EN Demo

Glossary

Network flow matrix

A network flow matrix is a table of the network communications a system needs: the source, destination, protocol, port, direction and justification of each flow. It is the reference firewall rules are written from, and the reference they are later reviewed against.

Updated

What is a network flow matrix?

A network flow matrix is a table of the network communications a system needs in order to work: which source talks to which destination, over which protocol and port, and why. You will also see it called a communication matrix or, when it is written for the security team, a firewall flow matrix.

The matrix states a need. It does not describe a configuration. One row says that the payroll web server queries its database over TCP 5432, in production, because the application reads and writes payslips there. How the firewall permits that is a separate question. Keeping the two apart is what makes the matrix useful: rules are written from it, and later checked against it.

It shows up at three points in a system's life. At design time it sits in the architecture document next to the network diagrams. At go-live it becomes the firewall change request. In operations it is what a firewall rule review compares the rule base to.

What columns should a flow matrix template have?

Eight columns cover most cases. A flow matrix template with thirty tends to come back half empty.

Source and destination name things you can identify: a server, a server group, a subnet, a zone. "Any" is not a source. Protocol and port say what is carried: TCP 443, UDP 53. Direction says which side opens the connection, something an arrow on a diagram does not always settle. Justification is one sentence in business terms ("posts accounting entries to the ERP") that someone joining six months later still understands. Owner is the person who can answer for the flow, usually whoever is responsible for the calling application. Environment separates production, test and development, because a flow opened for a test campaign has no business surviving in production.

A sample row: source "payroll-web (application DMZ)", destination "payroll-db (data zone)", protocol TCP, port 5432, direction web to database, justification "payroll application reads and writes payslips", owner "payroll application manager", environment "production".

Two more columns are worth adding: the change request reference, and an end date for temporary flows. NIST SP 800-41 says rulesets should be documented with comments on each rule wherever practical. The French cybersecurity agency ANSSI lists what such a comment can carry: a functional description, the date and the ticket reference. A matrix and a rule base that share references are far easier to reconcile. A flow matrix template in Excel format is available to download: sixteen columns, with the eight essential ones flagged in the "How to use" tab.

Who owns the flow matrix, and where does it live?

Three roles touch the matrix, which is often why nobody maintains it. The architect or project lead writes it during design. The application owner knows what the application really needs. The network or security team turns it into rules. If the matrix stays inside the project's design document, it stops moving on go-live day while the application keeps changing.

What holds up over time is one matrix per application or per scope, owned on the application side, and changed through the same request that opens or closes the rule. The architecture document keeps the version approved at project time, in its network chapter beside the zone diagram. The living version is kept somewhere operations actually looks, using the same application, server and zone names as the map of the IT estate. When names differ, matching flows to rules becomes manual work.

What is the difference between a flow matrix and firewall rules?

The matrix lists required flows. The firewall policy lists opened flows. They should overlap, and the gap between them falls into a few cases.

A flow is covered when a rule permits it and opens nothing more. It is missing when the matrix asks for it and no rule lets it through. A rule is unjustified when it permits traffic that no row of the matrix asks for. A fourth case deserves its own name: the rule exists but does not match the need, because it opens a whole subnet where one server was enough, or a port other than the one in the matrix.

Drift nearly always comes from the rule side. ANSSI's note on cleaning up a firewall policy lists causes of losing control: a policy that can reach several hundred or even thousands of rules, no writing conventions, frequent turnover among the people operating the firewalls, and a poor grasp of the map of the systems. In day-to-day terms that means the rule opened during a Friday incident and never closed, the temporary rule from a hypervisor migration still there after cutover, and the decommissioned server whose address was reassigned while its rules stayed.

How do you run a firewall rule review against the flow matrix?

The starting principle is deny by default. NIST SP 800-41 says firewalls should block all inbound and outbound traffic that has not been expressly permitted, and ANSSI's model rests on the same rule. NIST also asks that ruleset changes go through formal change management and that rulesets be reviewed periodically. ANSSI sets a frequency: a filtering policy should be reviewed "twice a year, or once a year at a minimum" (recommendation R15, our translation).

A review runs in this order. Export the rules of each firewall and the matrix into comparable formats, with the same object names or a lookup table between them. Start from the matrix: every required flow should find its rule. Then start from the rules: every rule should find its row, and the ones that do not go to their presumed owner with a closed question ("is this flow still needed?"). Rules that are too broad get narrowed to the written need.

An unjustified rule that still carries traffic goes to its presumed owner. For rules whose use is unknown, ANSSI's cleanup method is cautious, with good reason: turn on logging for the rules under review, analyze the logs over a representative period, go through the ones the logs show as unused with the people concerned, then disable them before deleting them. A rule that sees no traffic for a month may still serve the annual accounting close.

A review corrects the matrix as much as the rule base. A real, legitimate flow that was missing from the matrix gets a row and an owner.

Which tools can maintain a flow matrix?

The spreadsheet is still the most common tool: everyone can fill it in, it filters well, and it attaches to a design document. Nothing ties it to the rules, and two copies diverge quickly.

A CMDB can hold flows as relationships between configuration items. It gives every item an owner. In return, someone has to keep those relationships current (what a CMDB is and what it contains).

Firewall policy management tools, from firewall vendors or specialized software vendors, analyze the rules themselves: redundant rules, duplicate or unused objects, rules with no traffic. ANSSI recommends automated analysis tools for cleanup work (R1). These tools know the rules very well and the application need less well.

Traffic observation (firewall logs, NetFlow, probes) shows what actually flows. It helps rebuild a matrix that was never written, with one caveat: an observed flow is not thereby a justified one, and a rare flow can fall outside the observation window.

The last approach is a map derived from the sources: read the existing matrix, the rule exports and the inventory, then match them by address and name. It is only as good as the files it reads, and it leaves the decision to the people who know the application.

What Uncia brings, and what it does not do

Uncia reads, read-only, your flow matrices and the rules of your firewalls and clouds, through a connector or a dropped-in file, and reconciles them with the rest of the inventory. It checks the matrix of required flows against the actual rules and returns a verdict per flow: covered, missing, inconsistent, unjustified. Each verdict says why, rule in hand, for instance when a rule allows the flow more broadly than needed. The generated network map shows zones, the firewalls between them and the flows crossing them, and the thread runs from business capability to firewall rule in both directions. Every piece of information cites its source.

The limits are clear. Uncia changes no rule and administers no firewall. It does not observe traffic: it compares what is written on one side with what is written on the other, and the result updates with each collection. It does not decide that a flow is legitimate: it flags the rule with no justification, and a person makes the call. Uncia also produces reports structured along NIS2, DORA or ISO 27001; a structured report is not the same thing as compliance.

Sources

Frequently asked questions

How often should firewall rules be reviewed?

NIST SP 800-41 Rev. 1 asks for periodic ruleset reviews without setting a frequency. ANSSI, the French cybersecurity agency, recommends reviewing a filtering policy twice a year, or once a year at a minimum (technical note DAT-NT-006, R15). Reviews go faster when every rule carries the reference of the request that created it.

Should there be one flow matrix per application or one for the whole estate?

One matrix per application or per scope is easier to keep current, because it has an owner who knows the need. A consolidated view is still useful to the network team, provided it is built from the per-application matrices instead of being maintained in parallel.

Can a flow matrix be rebuilt from existing firewall rules?

That gives you the list of opened flows, which says nothing about required flows. It is a starting point: each rule then has to be confirmed by the owner of the application concerned, and rules nobody claims are examined separately.

What should you do with a firewall rule that has no justification?

If it carries traffic, the question goes to its presumed owner. If the logs show it as unused, ANSSI's cleanup method (technical note DAT-NT-32) works in steps: turn on logging for the rule, analyze the logs over a representative period, get agreement from the people concerned, disable the rule, then delete it after a long enough validation period.

Where does the flow matrix go in an architecture document?

In the network chapter, next to the diagram of zones and firewalls, or in an appendix when it is long. The document keeps the version approved at project time; the version used in operations lives in a separate repository.

Every week, an article to know your IT estate better

Architecture documents, end of support, NIS2, CMDB, migrations: methods you can apply with or without Uncia.

Get a call from our team

Thirty minutes with a member of our team, remote or on site: Uncia on a demonstration instance, then on part of your own estate if you wish. We reply within 48 hours.

Request received

Thank you. A member of our team will reply within 48 hours to set a date.