When designing network authentication, FortiAuthenticator can act as a RADIUS AAA server for network devices, but a secure design depends on much more than entering a shared secret and pointing a switch or firewall at the appliance. RADIUS requests arrive from registered clients, FortiAuthenticator selects a matching policy, that policy determines the authentication type and identity source, and the response can carry authorization attributes that influence network access. Each stage is a trust decision.
That is why RADIUS architecture should be designed around explicit boundaries: which devices may ask for authentication, which users or endpoints are eligible, which method proves identity, which attributes are trusted, how authorization is expressed, and what happens when certificates, directories, network paths, or time services fail.
FortiAuthenticator provides an identity-enforcement layer inside Fortinet security operations. If its policies are too broad, attackers may obtain access through the path of least resistance. If they are too brittle, routine infrastructure failures become authentication outages.
Register clients as trust anchors
FortiAuthenticator does not accept arbitrary RADIUS requests by default. A network device must be configured as a RADIUS client, identified by source address and a shared secret, and assigned to policy. This makes the client registry the first trust boundary. Treat it with the same care as a list of privileged infrastructure.
Use stable management addresses, restrict network paths to the RADIUS service, rotate secrets, and avoid reusing one secret across large groups of unrelated devices. A leaked secret should not allow an attacker to impersonate every access switch, VPN gateway, and wireless controller.
Client naming should also support operations. Include site or function information so logs identify which device generated the request. When a user reports an authentication failure, the RADIUS client should be obvious enough that the support team can distinguish a local switch problem from a policy or identity-source problem.
RADIUS itself was designed in an era when network trust assumptions were different, so protect the protocol with network segmentation and modern transport boundaries. Limit which interfaces accept requests and which devices can reach them. If RADIUS crosses untrusted networks, use architecture that provides additional protection rather than assuming the shared secret encrypts the entire exchange.
Treat client onboarding as a change-controlled process. A new switch, VPN concentrator, or wireless controller should have a named owner, documented secret exchange, intended authentication method, and test plan before it is added to production policy.
Policy matching must be unambiguous
FortiAuthenticator evaluates RADIUS policies against the requesting client, optional RADIUS attribute criteria, and the authentication type. If multiple policies match, order matters: the policy closer to the top of the list is selected. That makes policy order part of the security model.
Design policies so that one request has one intended path. Broad catch-all policies placed above specific EAP-TLS, MAB, or privileged-access policies can silently defeat the more restrictive logic. During review, test representative requests against the full ordered list rather than evaluating policies one at a time.
This resembles the broader problem of identity architecture: the control is only as strong as the path selection logic that decides which proof is acceptable. Ambiguous matching turns a carefully designed strong method into an optional branch.
Policy ordering should be reviewed whenever a new policy is added. The safest procedure is to test representative requests for every nearby policy, not only the new one. A new broad policy can alter the result of older requests even though those older policies were not edited.
Use descriptive policy names that reveal method and scope, such as site, client class, or authentication type. This reduces the risk that operators reorder policies based on vague names and unintentionally change access behavior.
Choose authentication methods for the actual endpoint
FortiAuthenticator policies can support password/OTP, MAC authentication bypass, and EAP-TLS. These methods solve different problems. EAP-TLS provides certificate-based mutual authentication suitable for managed users and devices. Password and OTP can support interactive access where certificates are unavailable. MAB can accommodate devices that cannot perform 802.1X, but a MAC address is not a strong identity proof.
For managed enterprise endpoints, 802.1X authentication with EAP-TLS generally creates a stronger trust model than password-only access because the device or user must possess certificate material issued by a trusted authority. Certificate lifecycle therefore becomes part of availability: enrollment, renewal, revocation, trust-chain changes, and clock accuracy all need operating procedures.
MAB should be isolated into narrower authorization outcomes. Printers, phones, sensors, and legacy devices may need connectivity, but they should not inherit the same network privileges as a strongly authenticated workstation merely because both ultimately receive Access-Accept.
Authorization is where authentication becomes network policy
A successful RADIUS exchange can return attributes that influence VLAN assignment, roles, access-control behavior, session parameters, or other device-specific outcomes. Those attributes should be treated as security-sensitive output, not as convenience metadata.
Coordinate RADIUS responses with the organization’s network access control design. The policy should answer what an authenticated identity is allowed to reach, not just whether it may connect. Segmentation, role mapping, guest treatment, contractor access, and remediation networks all depend on consistent attribute semantics between FortiAuthenticator and the enforcement device.
Document which attributes each client understands and how it behaves when an attribute is missing or malformed. A switch and a VPN gateway may interpret the same response differently. Testing should include not only successful authentication but also the exact authorization result on the network.
Authorization responses should be validated on the enforcement device, because vendors differ in attribute support and precedence. Capture an accepted exchange during testing and verify the endpoint actually lands in the intended VLAN, role, or policy. An Access-Accept with an ignored attribute is a security failure disguised as authentication success.
Plan for authorization rollback. If a new RADIUS response causes unexpected segmentation, operators should know how to revert to a safe known state without disabling strong authentication for the entire environment.
Certificates, directories, and time are hidden dependencies
EAP-TLS depends on certificate trust and valid time. Password and OTP policies may depend on local or remote identity stores. Remote LDAP or RADIUS dependencies add DNS, routing, firewall, and service availability to the authentication chain. A design that looks simple in the FortiAuthenticator GUI may therefore rely on several external systems.
Map those dependencies and define failure behavior. If a directory is unreachable, should access fail closed? Is there a local emergency path for administrators? If the certificate authority changes intermediates, which endpoints and appliances need updated trust? If NTP is wrong, which certificate validations or OTP windows break first?
Strong authentication does not eliminate operational risk. Even phishing-resistant authentication can fail at surrounding architecture boundaries such as enrollment, recovery, help-desk override, or mis-scoped authorization.
Certificate monitoring should cover both server identity and client-certificate ecosystems. Expiring issuing CAs, revoked intermediates, changed EKUs, or new trust anchors can affect thousands of endpoints at once. Test renewal procedures before the production chain approaches expiry.
Directory latency also matters. Authentication can fail because an upstream LDAP or identity service is slow rather than unavailable. Monitor response time separately from outright errors so capacity problems appear before they become login outages.
Troubleshoot by following the request path
When authentication fails, work from the request inward. Confirm that the network device can reach FortiAuthenticator on the expected interface, that the source address matches a registered client, that the shared secret is correct, and that the service is enabled on the interface. Then determine which policy matched and whether its authentication type aligns with the request.
Next inspect the identity source, certificate chain or password/OTP validation, group membership, and RADIUS response. Logging changes to clients and policies is useful because configuration edits can restart or alter RADIUS behavior. Correlate the user complaint with configuration history before assuming the endpoint is at fault.
Capture enough evidence to distinguish reject, timeout, and authorization problems. A timeout indicates a different class of fault from an explicit Access-Reject, while successful authentication followed by no network access usually points to response attributes or enforcement policy.
Design for controlled failure and recovery
FortiAuthenticator should be operated as a critical identity service inside the broader Fortinet platform. Backups, high availability where appropriate, tested restore procedures, documented break-glass access, secret rotation, certificate-renewal monitoring, and policy-review cadence all contribute to security.
Measure authentication success rate by client and method, reject causes, timeout rate, certificate-expiration horizon, policy-match distribution, MAB usage, and emergency overrides. These metrics can expose a migration problem before it becomes an outage or show where weak methods remain more common than intended.
Most importantly, do not treat RADIUS success as the end of the control. The secure outcome is a verified identity receiving the minimum network authorization appropriate to its role, with enough logging to explain the decision later.
Use synthetic authentication tests from representative network locations. A scheduled EAP-TLS and password/OTP test can detect routing, certificate, or policy regressions before users report them. Keep the synthetic identity tightly controlled and clearly marked in logs.
Periodic access reviews should also challenge whether MAB exceptions are still necessary. Devices that once lacked 802.1X support may have been replaced, yet the weaker policy often remains indefinitely unless someone owns its retirement.