Network segmentation

On this page 7

Network segmentation is the practice of dividing a network into separate zones and controlling which traffic may pass between them according to explicit policy. In network security, it limits the blast radius of an intrusion, a misconfiguration or malware, because a compromised device can reach only what its own segment is allowed to reach.

Segmentation combines two ingredients: a way to separate traffic, such as VLANs, subnets, separate physical networks or virtual networks, and an enforcement point that filters what crosses the boundary, such as a firewall, a router with access control lists or a host-based policy engine. Separation without enforcement is only organization, not security.

How network segmentation enforces boundaries

Segmentation works by forcing traffic between zones through a policy decision point. Devices in the same segment usually communicate directly at Layer 2, while traffic to another segment must be routed, and the router or firewall applies rules based on source, destination, protocol and port.

A well-designed segment groups systems with similar trust level and function, for example user workstations, servers, management interfaces, guest devices or industrial controllers. The rules between zones follow the principle of default deny: only flows with a documented business reason are permitted.

The table below lists the technical building blocks and the network layer at which each one operates.

MechanismLayerRole in segmentation
VLAN (IEEE 802.1Q)Layer 2Separates broadcast domains on shared switches
IP subnetLayer 3Gives each zone its own address range for routing and rules
Stateful firewallLayers 3-4, sometimes 7Enforces policy between zones and tracks connections
Network ACLLayers 3-4Stateless filtering on routers, switches or cloud subnets
VRFLayer 3Separate routing tables on one router
Host firewall or agentHostEnforces policy per workload for microsegmentation
Building blocks of segmentation

Types and granularity of segmentation

Segmentation ranges from a few coarse zones to policies around individual workloads. Finer granularity gives stronger containment but requires better visibility and more rule maintenance.

ApproachTypical boundaryStrengthEffort
Physical separationSeparate cabling and switchesVery strong isolationHigh hardware cost
MacrosegmentationVLANs and subnets per zoneGood containment between zonesModerate
MicrosegmentationPer application or per workloadLimits lateral movement inside zonesHigh, needs flow visibility
Identity-based segmentationUser or device identity via 802.1X and RADIUSPolicy follows the deviceHigh, needs identity infrastructure
Levels of segmentation

Many organizations start with macrosegmentation and add microsegmentation for the most sensitive systems, such as payment processing, domain controllers or backup infrastructure. Industrial environments often apply a zones and conduits model from IEC 62443, which treats each permitted path between zones as a controlled conduit.

Segmentation is often confused with neighboring controls that serve different purposes. A VPN, for instance, protects traffic in transit, but by itself it does not decide which internal systems a connected user may reach.

ControlMain purposeRelation to segmentation
Network segmentationRestrict reachability between zonesThe architecture itself
Network ACLFilter packets by address and portOne enforcement tool inside the design
VPN, IPsec or WireGuardEncrypt and authenticate tunnelsShould land users in a restricted segment
RADIUS with 802.1XAuthenticate devices at the portAssigns devices to the correct VLAN
Zero trust accessVerify every request by identityExtends segmentation to the application level
Segmentation and neighboring controls

In practice these controls reinforce each other. A remote user arrives through a VPN, lands in a dedicated remote-access segment, and reaches only the applications that the firewall policy allows for that user group.

Designing and maintaining segments

Effective segmentation begins with an inventory of assets and data flows, because rules cannot be written for traffic nobody has mapped. Flow data from NetFlow or firewall logs collected over several weeks reveals the dependencies that a design must preserve.

The decision table below matches common situations with a reasonable segmentation choice.

SituationRecommended choiceReason
Guest Wi-FiSeparate VLAN with internet access onlyUntrusted devices never reach internal systems
Management interfacesDedicated out-of-band or management VLANProtects switches, hypervisors and controllers
Cardholder dataIsolated zone with strict rulesReduces PCI DSS scope
IoT and camerasOwn VLAN, outbound to required services onlyDevices are rarely patched
Backup serversRestricted zone, pull-based where possibleHinders ransomware from reaching backups
Segmentation decisions

Rules must be reviewed regularly, because temporary exceptions tend to become permanent. A useful habit is to attach an owner and an expiry date to every rule that permits traffic between zones.

Limitations and common mistakes

Segmentation reduces risk but does not stop attacks that use permitted paths. An attacker who compromises a web server can still reach the database the web server is allowed to query.

  • Flat VLANs with any-any rules: zones exist on paper, but the firewall allows everything between them.
  • Forgotten management paths: jump hosts, monitoring servers or backup agents often have broad access and become the bridge an attacker needs.
  • Shared credentials across zones: identical administrator passwords defeat network boundaries.
  • No monitoring at boundaries: denied flows are a valuable early signal of lateral movement and should be logged.
  • Over-segmentation without tooling: thousands of hand-written rules become impossible to audit.

In simple terms

Picture a ship with watertight compartments. If the hull is damaged in one place, water fills only that compartment, and the ship stays afloat. Network segmentation applies the same idea to computers.

Instead of one large open network where every device can talk to every other device, the network is divided into rooms with locked doors. Guards at each door let through only the visitors who have a reason to pass. If an intruder gets into one room, the rest of the building stays protected.

A practical example

A regional clinic runs staff computers, medical devices, a patient records server, security cameras and guest Wi-Fi on one flat network. A risk assessment shows that malware on any reception computer could reach the records server directly.

The network team creates six VLANs: staff, servers, medical devices, cameras, management and guests. A firewall between them allows staff computers to reach the records application only on TCP 443, lets medical devices talk only to their vendor gateway, and gives guests internet access only. All denied flows are logged.

Months later, a staff laptop is infected through a malicious email attachment. The malware tries to scan the server subnet, but the firewall blocks and logs the attempts. The security team isolates the laptop within an hour, and no server or medical device is affected.