In many businesses, IP cameras run on the same flat network as computers, printers and servers. Installation is easy: the camera is plugged in, gets an address from DHCP and the NVR finds it. The problem comes later: an unpatched camera becomes a door into the corporate network, multicast streams spill over to office ports, and after an incident nobody can answer “whose device is this?”. Below is why the camera network should be separated, what segment model to build, common configuration mistakes and how to migrate without losing recordings. To fit the camera segment into the company’s overall VLAN plan, see also the VLAN design in a corporate network guide.
Why a separate camera network?
Security
Cameras are embedded computers running on the network. There are many of them, they sit in physically accessible places and they’re often not updated as frequently as corporate computers. Vulnerabilities affecting camera and NVR products regularly appear in lists of exploited vulnerabilities. On a flat network, a single compromised camera can be a stepping stone for an attacker to reach corporate systems directly. Segmentation limits this spread.
There’s also a physical access dimension. Someone who unplugs the cable of a camera on a car park pole and connects a laptop can reach all the way to the server VLAN on a flat network. A separate segment with rule-based traffic ensures that port can only talk to the recording server.
Performance
Video traffic is continuous and high. A system where 50 cameras produce an average of 3 Mbps carries 150 Mbps constantly. If multicast streams aren’t managed correctly, they can spill over to every port on the network. A separate segment and suitable QoS/IGMP settings keep video traffic from affecting business applications.
Management and audit
A separate address block makes inventory, monitoring and incident investigation easier. When frameworks such as TISAX and ISO 27001 ask for evidence of network separation and access control, a documented segment structure can be used directly. When the auditor asks “can the cameras reach the corporate network?”, the answer is a rule list. Not a promise.
Recommended segment model
| Segment | Content | Example address |
|---|---|---|
| Camera VLAN | IP cameras, encoders | VLAN 30 · 10.30.0.0/22 |
| Recording segment | NVRs, VMS recording and management servers | VLAN 31 · 10.31.0.0/24 |
| Client segment | Security operator workstations, video wall | VLAN 32 · 10.32.0.0/24 |
| Management network | Management access to switches, cameras and NVRs | VLAN 99 · separate and restricted |
In small systems, cameras and recording can be in the same segment; what matters is that cameras are separated from the corporate user network and the internet, and that traffic between them is governed by rules. In production facilities, the camera network should also be separated from the OT (production control) network; NIST SP 800-82 recommends zoning and controlled conduits in OT environments.
Choose the address block with room to grow. The /22 in the example gives space for about a thousand devices; at a facility with 60 cameras today this may look excessive, but re-addressing the camera VLAN a second time is more work than the first. Reserving sub-blocks by site or building also simplifies inventory and ACL writing.
Flow matrix: who can talk to whom?
- Cameras → recording server: only the streaming and management ports used by the VMS
- Cameras → NTP server: time synchronization
- Clients → VMS server: VMS client ports (clients don’t connect to cameras directly)
- Management network → cameras, NVRs, switches: HTTPS and SSH, authorized administrators only
- Cameras → internet: closed by default; controlled access to specific update servers if needed
- Corporate user network → camera network: closed
Port numbers are taken from the official documentation of the VMS and devices in use and written into the matrix with their justification. This matrix is the main output of our segmentation project.
Rules are ordered as “permits first, an explicit deny at the end”, and denied traffic is logged. In the first week, these logs show dependencies forgotten in the matrix (such as DNS, syslog or a license server). A deny rule without logging makes failures silent.
Multicast and IGMP
Some VMS products use multicast for live viewing. If IGMP snooping isn’t enabled on the switches, multicast traffic is sent to all ports. IGMP snooping should be enabled in the camera VLAN and there should be an IGMP querier in the segment. Even if multicast isn’t used, these settings are still useful for limiting unexpected traffic.
A missing querier is a common error in the field that’s noticed late. If snooping is on but there’s no querier, switches forget group memberships after a while; live viewing works for a few minutes and then freezes, or multicast spills back to all ports. If there’s no multicast router in the segment, give the querier role to an L3 switch or the core switch.
Common configuration mistakes in the field
- VLAN created, no ACL written: An SVI is created for the camera VLAN on the core L3 switch, and inter-VLAN routing is open by default. The separation exists only in the address plan.
- All VLANs allowed on the trunk: Only the necessary VLANs should be allowed on the uplink trunk to the camera switch; otherwise the camera switch also carries corporate VLANs.
- Camera port not in access mode: Camera ports are configured in access mode in the relevant VLAN; unused ports are shut down or moved to an empty VLAN.
- NVR connected to two networks: When the LAN port of an NVR with PoE ports is plugged into the corporate network, the device ends up bridging the two networks. The NVR’s interface facing the corporate side must also be in a rule-governed segment.
- Management interface open from everywhere: If the camera web interface is reachable from the user network, the matrix hasn’t been applied. Management is done only from the management network.
On Cisco IOS-based switches, allowed VLANs on a trunk are checked with show interfaces trunk and IGMP snooping status with show ip igmp snooping; for other vendors, the equivalent commands are in their own documentation.
Remote access: instead of port forwarding
Giving access to the NVR from the internet via port forwarding exposes the device to automated scans. The right method is a VPN or an intermediary service whose security features have been assessed, with personal user accounts and multi-factor authentication where possible (secure remote access). Mobile viewing with a shared “admin” account leaves who looked when unrecorded.
Steps for an uninterrupted migration
- Inventory: IP, MAC, switch port and dependencies of all cameras, NVRs and clients.
- Target IP plan and flow matrix; preparing firewall rules in advance.
- Preparing the recording server’s access to the new segment (two interfaces if needed).
- Moving cameras to the new VLAN group by group, verifying recording after each group.
- Closing old access and verifying the rules with a scan.
The item that wastes the most time in migration is cameras with static IPs: when the VLAN changes, the old address is invalid in the new network and you may need to go on site to reach the camera. Before moving a group, plan the address change on the camera side and update the device entries in the VMS in the same window. After each group, check not just the live image but that recording is actually being written.
Frequently asked questions
Does a VLAN provide security on its own?
No. A VLAN provides logical separation; security comes from firewall or ACL rules that limit traffic between VLANs. If routing between VLANs is open without rules, the separation stays on paper.
Can cameras use DHCP?
They can, but DHCP reservations or static addressing are preferred so an address change doesn’t break the recording connection; the IP–device mapping is kept in the inventory.
Do we need to open ports for mobile viewing?
No. Instead of port forwarding, use a VPN or an intermediary access service whose security features have been assessed.
Segmentation makes sense when handled together with device hardening; for the checklist items, see our IP camera and NVR cybersecurity checklist.