Unplanned Maintenance

Networking

VLANs separate networks. Your policy decides what crosses.

Follow the traffic from a camera to its recorder to see where a VLAN ends and firewall policy begins.

Sources checked . See the article’s scope for verification limits.

Putting a camera on an “IoT VLAN” is a useful start. It does not answer whether that camera can reach your storage server, your laptop or the switch's management page. The label on the diagram has no enforcement powers.

A virtual LAN (VLAN) divides a switched network into separate broadcast domains. A trunk can carry multiple VLANs over one link. Those are segmentation mechanisms; Netgate describes them in its VLAN overview. The security question comes next: which paths connect those segments, and what policy applies there?

This article develops a small illustrative design, with pfSense software as the concrete firewall example. It is a design and verification workflow, not a configuration to paste into a live router. You need administrator access, a VLAN-capable managed switch, compatible firewall interfaces, a port map, configuration backups and a management recovery path before making changes. If wireless clients are in scope, confirm that the access-point model and configuration support the intended VLAN attachment before including them.

Draw the forwarding path before the rules

For the example, put clients in VLAN 10, services in VLAN 20 and cameras in VLAN 30. Use a distinct IP subnet for each. Place their gateways on the firewall so that intended traffic between these subnets crosses it.

That last choice matters. Devices in the same VLAN can communicate directly without routing; a Layer 3 switch can also route between VLANs. If such a switch provides both gateways, traffic between those networks need not visit an upstream Internet firewall. Cisco's inter-VLAN routing guide describes these paths.

The resulting design rule is ours: identify the actual enforcement point for each important flow. If you deliberately route on a switch, evaluate the switch's access controls and their capabilities. Do not assess protection from rules on a firewall outside that path. Likewise, do not expect a routed firewall to isolate two ordinary hosts sharing the same switched VLAN; that requires a boundary or controls appropriate to their local path.

Write a traffic policy in plain language

In this hypothetical environment, a designated administration laptop manages the infrastructure. A recorder initiates a specific supported connection to each camera. The Domain Name System (DNS) resolver and time service have known addresses. Here, HTTPS means HTTP over TLS, and SSH means Secure Shell. Ports below are example policy choices, not a claim about every camera or recorder.

SourceDestinationServiceIntended result
Admin laptopManagement addressesSelected HTTPS/SSH administration portsAllow
Ordinary clientsInventory web applicationTCP 443Allow
CamerasDesignated DNS resolverUDP/TCP 53Allow
CamerasDesignated time serviceUDP 123Allow
RecorderSpecific camerasDocumented recorder-initiated serviceAllow after checking vendor requirements
CamerasOther internal addressesOther new connectionsDeny
CamerasInternetNew connectionsDeny unless a separately justified exception exists

The service-number examples follow IANA registrations for DNS, Network Time Protocol (NTP) and HTTPS. A port number alone does not establish which application is using it.

This is a restrictive starting policy. It may be unsuitable for cloud-dependent devices or cameras that initiate the recording connection. Before implementation, check the actual device documentation and existing traffic requirements. Resolve those differences explicitly rather than adding a broad allow rule because recording stopped.

Also account for addressing services such as DHCP (Dynamic Host Configuration Protocol) if used, and for services hosted on the firewall itself. This table describes application intent, not a complete ruleset for booting and operating every client.

Translate the policy into this firewall's behavior

For ordinary pfSense interface-tab rules, filtering occurs where the initiating traffic enters the firewall. Camera-originated traffic therefore belongs on the camera interface. Floating and interface-group rules can affect processing before interface rules, so review the effective ruleset rather than one tab in isolation. These details are specific to pfSense rule methodology; other products can use different models.

Ordinary interface-tab rules use first-match ordering. A broad allow above a restriction can make the restriction ineffective. Stateful filtering also permits replies to an allowed connection. In our design, permitting the recorder's connection does not require a broad camera-originated allow for its replies. These behaviors are documented in pfSense firewall fundamentals.

Review IPv4 and IPv6 deliberately. A pfSense rule matches only its selected address family. An IPv4 restriction is not evidence that the corresponding IPv6 path has been restricted. Specify policy and verification for both families where enabled, using the rule configuration reference. Do not invent an IPv6 policy by indiscriminately blocking all control traffic; use the installed product's documented IPv6 requirements.

Keep the switch configuration in the review

The firewall cannot repair an incorrectly assigned switch port. Check which ports are access ports for end devices and which trunks are allowed to carry the required VLANs. Document tagging and native/untagged behavior at both ends of each trunk.

Restrict trunk access and avoid casually using a default VLAN for sensitive management. Netgate's VLAN security guidance explains native-VLAN and trunk-exposure concerns. That guidance also treats substantially different trust zones as a reason to consider physical separation. VLANs are a useful tool, not a universal answer to every isolation requirement.

In pfSense, creating a VLAN alone does not finish its setup: the interface must also be assigned, enabled and configured, with rules and required services. See VLAN configuration. Treat the switch, firewall and client attachment as one end-to-end change.

Stage one client and test the boundaries

Our recommended rollout starts with a spare access port and one noncritical test device. Keep the existing administration path intact. Save the current firewall and switch configurations, record the original port membership and confirm local console access before editing management paths.

For each policy row, record source, destination, protocol/port, address family, expected result and observed result. Use a real listening service for a deny test; otherwise a failed connection might only demonstrate that nothing was listening. Only probe systems you administer.

The first checks should establish the test device's address, gateway and required name resolution. Then verify a permitted application connection and attempt a specifically forbidden connection. Check the firewall's observations against the expected path. A single successful ping is not proof that either the application policy or the full isolation policy works.

Use new connections when validating changes. pfSense maintains connection state; its state-table interface lets an administrator inspect and remove individual states. If a test needs state removal, target only the relevant test connection after checking its identity. Clearing the entire table can interrupt unrelated work and should not be a casual troubleshooting step.

Test from each relevant source segment and for each enabled address family. Confirm that ordinary clients cannot use the administration exception. Expand the rollout only when observed behavior matches the written policy.

Have a way to undo the experiment

Stop if a forbidden connection succeeds, management access becomes unreliable or a required service fails without an understood cause. Restore the test port's original membership and the affected rules using the saved plan, then recheck the prior behavior. Avoid moving the only management connection while simultaneously changing its VLAN and access policy.

If broader configuration restoration is needed, use the device's supported recovery procedure. For pfSense, configuration revision compatibility matters; a full GUI restore can reboot the firewall. Read Netgate's restore guidance before treating configuration import as a harmless undo button.

The useful outcome is a documented boundary with observable behavior: these sources can reach these services, other intended paths are denied, and management still works. A spreadsheet of VLAN numbers is supporting paperwork. The traffic decides whether the design is finished.

Sources and scope

Primary documentation checked 2026-11-17. This is a hypothetical design with documentary verification, not a tested deployment or complete security assessment. Match product versions, device service requirements and local recovery arrangements before use.

All articles