In most organizations, anyone who plugs a laptop into the empty socket in the meeting room lands directly in the user VLAN. Nobody asks, nobody records it. 802.1X-based network access control (NAC) verifies who the device connecting to the port is and gives it the appropriate network; but 802.1X switched to enforcement mode overnight knocks printers, IP phones and clients missing certificates off the network in the morning. The plan below is for rolling out 802.1X on the wired network in measurable phases while keeping a way back open.
What NAC solves and what it doesn’t
802.1X has three roles: the connecting device (supplicant), the switch or wireless access point (authenticator), and the authentication server (RADIUS). Until authentication completes, the switch lets only EAPOL frames pass through the port; the RADIUS server makes the decision and, if needed, returns a VLAN, a downloadable ACL (dACL) or other authorization attributes. Using 802.1X together with RADIUS attributes is described in the informational RFC 3580.
The problems NAC solves: unauthorized devices’ direct access to the internal network, assigning different networks and permissions by device type, and recording which device connected to which port and when. What it doesn’t solve is at least as important from an enterprise cybersecurity perspective. It doesn’t stop what an authenticated but compromised computer does, it doesn’t replace segmentation, and it’s weak against MAC spoofing for devices admitted with MAB. The segmentation boundaries described in the difference between VLANs and real network security article still apply after NAC. NAC is only the guard at the door. It doesn’t see what happens inside.
Authentication methods: EAP-TLS, PEAP and MAB
| Method | Proof of identity | Strength | Limitation |
|---|---|---|---|
| EAP-TLS | Client certificate | No password; mutual certificate authentication | Requires certificate distribution and renewal infrastructure |
| PEAP-MSCHAPv2 | User/machine password | Easy start, works directly with AD | Credentials can be captured if the server certificate isn’t validated |
| MAB | MAC address | The only option for devices without a supplicant | Can be bypassed with MAC spoofing; it’s identification, not authentication |
For corporate Windows clients, the goal is EAP-TLS with a machine certificate (RFC 5216). Machine authentication lets the device get onto the network before the user signs in; group policy, logon scripts and remote management depend on this. PEAP can be used as a migration step. In that case, validating the RADIUS server certificate and defining trusted server names must be mandatory on clients; otherwise a rogue access point or switch can collect password hashes. This is the common PEAP mistake: a profile with the “Verify the server’s identity” box left unchecked.
RADIUS and certificate infrastructure design
The moment 802.1X is turned on, RADIUS becomes critical infrastructure: if the server doesn’t respond, newly connecting or re-authenticating devices can’t get onto the network. Settle these decisions in advance in the design:
- Redundancy: At least two RADIUS nodes, preferably in different locations; two servers and a dead server detection (dead criteria) definition on the switches. A single RADIUS server is a single point of failure.
- Critical VLAN / unreachable server behavior: Which network the port is placed in if all RADIUS servers are unreachable is decided in advance.
- Identity source: Active Directory groups are the basis of authorization rules; device and user groups are kept separate.
- PKI: Autoenrollment for machine certificates, renewal, and the reachability of revocation (CRL/OCSP). If the RADIUS server can’t reach the CRL, it may reject all authentications; this can show up as a mass outage in the morning.
- Logging: RADIUS accept/reject records are sent to the SIEM. Reject reasons are the main source for troubleshooting.
Monitor, low-impact and closed mode phases
The essence of the migration is to run 802.1X first without blocking anything and collect data. The three-phase model defined by Cisco, with similar equivalents at other vendors, is as follows:
| Stage | Port behavior | Purpose |
|---|---|---|
| Monitor | Traffic passes even if authentication fails | Seeing which device fails and why |
| Low-impact | Before authentication, only traffic allowed by the pre-ACL (DHCP, DNS, AD, PXE) passes | Partial enforcement; full access via dACL for successful devices |
| Closed | Without authentication, only EAPOL passes | Full enforcement; the highest security |
Moving between phases is based on criteria, not dates. The typical criterion for leaving monitor mode is seeing a successful 802.1X or a deliberately defined MAB result on almost all active ports, with each remaining failing device explained. Do the migration building by building or floor by floor; if something goes wrong, the impact stays on that floor.
Strategy for devices that don’t support 802.1X
Printers, IP phones, badge access panels, IP cameras and production devices are the main source of failed authentications in most environments. A layered approach is followed for these devices:
- 802.1X if supported: Many enterprise printers and IP phones support 802.1X with a certificate or PEAP. Check this first.
- MAB + profiling: The MAC address is evaluated together with NAC device profiling data (DHCP, CDP/LLDP attributes), and the device is placed only in the VLAN and dACL for its own category.
- Narrow authorization: A printer admitted via MAB should be able to talk only to the print server, and a camera only to the recording server. That keeps the payoff of MAC spoofing limited.
- Computer behind an IP phone: Multi-domain authentication authenticates the phone in the voice VLAN and the computer behind it in the data VLAN separately.
If the ports of assets belonging to a separate network, such as cameras and industrial devices, are already in a separate segment, NAC rules get simpler. To standardize the access port template, the switch configuration standard guide on the network site is also useful.
Example switch-side configuration and verification
The example below shows an access port in monitor mode on Cisco IOS-XE using the classic command syntax. The authentication open line provides monitor mode; when moving to closed mode, this line is removed. The authentication event server dead lines determine which VLAN (the critical VLAN) the port is placed in when all RADIUS servers are unreachable; dead-criteria and deadtime define when a server is considered dead and for how long it’s skipped. On newer IOS-XE releases, the same configuration can be converted to the IBNS 2.0 policy syntax; verify the syntax suitable for your release in the vendor documentation.
aaa new-model
aaa authentication dot1x default group radius
aaa authorization network default group radius
aaa accounting dot1x default start-stop group radius
dot1x system-auth-control
!
radius server NAC-1
address ipv4 10.10.10.11 auth-port 1812 acct-port 1813
key <shared-secret>
!
radius server NAC-2
address ipv4 10.20.10.11 auth-port 1812 acct-port 1813
key <shared-secret>
!
radius-server dead-criteria time 5 tries 3
radius-server deadtime 15
!
interface GigabitEthernet1/0/10
switchport mode access
switchport access vlan 20
switchport voice vlan 30
authentication host-mode multi-domain
authentication order dot1x mab
authentication priority dot1x mab
authentication port-control auto
authentication open
authentication periodic
authentication timer reauthenticate server
authentication event server dead action authorize vlan 20
authentication event server dead action authorize voice
authentication event server alive action reinitialize
mab
dot1x pae authenticator
dot1x timeout tx-period 10
spanning-tree portfast
The first commands to check for verification and troubleshooting:
show authentication sessions interface GigabitEthernet1/0/10 details
show dot1x interface GigabitEthernet1/0/10 details
test aaa group radius <user> <password> new-code
debug dot1x all
debug radius authentication
Use debug commands briefly and carefully on busy switches; turn them off with undebug all when you’re done. Locking up the console session with debug output is a well-known mistake. For wired 802.1X on Windows clients, the “Wired AutoConfig” (dot3svc) service must be running; the client-side status is shown with netsh lan show interfaces.
Common authentication errors
- Supplicant off: The Wired AutoConfig service is disabled, or the wired 802.1X profile wasn’t distributed via group policy. The device falls back to MAB and is rejected as an unknown MAC.
- Certificate problems: There’s no machine certificate, it has expired, it isn’t in the CA chain the RADIUS server trusts, or the CRL can’t be reached.
- The server certificate isn’t trusted: The client doesn’t trust the CA that issued the RADIUS server’s certificate, or the server name doesn’t match the name in the profile.
- Timing problems: Devices that PXE boot or start up quickly can’t complete the DHCP request before the 802.1X timeout and the fallback to MAB expire.
tx-periodand the authentication order are tuned for these devices. - Machine/user transition: When the user signs in, re-authentication takes place; if the user’s authorization returns a different VLAN, the IP address doesn’t change and the connection drops. Authorization rules should be designed consistently for the machine and the user.
- Unmanaged switch or hub: In single-host mode a second MAC, and in multi-domain mode a second MAC in the data domain, triggers a security violation, and by default the port may go err-disabled; the behavior is set with
authentication violation. Inventory the desktop switches in meeting rooms. They’re all there.
To design NAC together with segmentation and Zero Trust goals, network security and segmentation and Zero Trust security architecture work can be handled together. The remote access counterpart of the same device posture approach is described in the ZTNA versus VPN comparison article.
Frequently asked questions
What happens if I turn on 802.1X directly in closed mode?
Clients whose supplicant isn’t configured, devices missing certificates, and devices such as printers and phones without a defined MAB rule can’t get onto the network. Moving to enforcement without seeing and fixing failed authentications in monitor mode causes a widespread outage.
Should I choose PEAP or EAP-TLS?
The long-term goal should be EAP-TLS; it uses no passwords and performs mutual certificate authentication. PEAP can be used for a quick start, but RADIUS server certificate validation must be mandatory on clients.
Is MAB a secure authentication method?
No, MAB is an identification method; MAC addresses are easily spoofed. Devices admitted via MAB should be backed by profiling data and limited with a narrow VLAN and ACL so they can reach only the systems they need.
If the RADIUS servers become unreachable, do users drop off the network?
Already authenticated sessions continue until re-authentication time. Newly connecting devices, depending on the unreachable server behavior, are either rejected or placed in a predefined critical VLAN. This behavior should be decided in the design and tested.