Passer au contenu principal

Migration de RADIUS sur site (NPS) vers RADIUS en tant que Service

Ce guide faisant autorité détaille l'architecture technique, la méthodologie de mise en œuvre et l'impact commercial de la migration du Microsoft Network Policy Server (NPS) sur site vers un modèle RADIUS en tant que Service natif du cloud. Il fournit aux leaders informatiques et aux architectes réseau des cadres pratiques pour réduire les frais opérationnels, éliminer les points de défaillance uniques et sécuriser l'authentification d'entreprise sur des sites distribués.

📖 5 min de lecture📝 1,066 mots🔧 2 exemples concrets3 questions d'entraînement📚 8 définitions clés

Écouter ce guide

Voir la transcription du podcast
PODCAST SCRIPT: Migrating from On-Premises RADIUS (NPS) to RADIUS as a Service Duration: ~10 minutes | Voice: UK English, Male, Senior Consultant tone --- SEGMENT 1: INTRODUCTION AND CONTEXT Welcome to the Purple WiFi technical briefing series. Today we're tackling a migration that's sitting on the roadmap of a significant number of enterprise IT teams right now: moving away from on-premises RADIUS — specifically Microsoft's Network Policy Server — to a cloud-hosted RADIUS as a Service model. If you're managing WiFi authentication across a hotel group, a retail estate, a stadium, or a public-sector campus, this is directly relevant to you. The on-premises NPS model has served us well for the better part of two decades, but the operational overhead, the single-point-of-failure risk, and the scaling limitations are becoming increasingly hard to justify — particularly when cloud-native alternatives now offer enterprise-grade reliability at a fraction of the total cost of ownership. Over the next ten minutes, we'll cover the technical architecture of both approaches, walk through a structured migration methodology, look at two real-world implementation scenarios, and finish with the key decision frameworks you need to make this call confidently. Let's get into it. --- SEGMENT 2: TECHNICAL DEEP-DIVE First, let's make sure we're aligned on what RADIUS actually does in your network stack. RADIUS — Remote Authentication Dial-In User Service — is the protocol defined in RFC 2865 that handles authentication, authorisation, and accounting for network access. In a WiFi context, it's the backbone of IEEE 802.1X port-based access control. When a device connects to a WPA2-Enterprise or WPA3-Enterprise SSID, the access point acts as a RADIUS client — what we call a Network Access Server — and forwards the authentication request to the RADIUS server. The server validates the credentials, typically against Active Directory or an LDAP directory, and returns an Access-Accept or Access-Reject response. That's the fundamental flow. Now, in the on-premises NPS model — Network Policy Server is Microsoft's RADIUS implementation bundled with Windows Server — you're running that authentication logic on hardware you own, in a data centre or server room you maintain. The NPS server holds your network policies, your certificate infrastructure for EAP-TLS or PEAP-MSCHAPv2, and your connection request policies. It works. It's mature. But it comes with a set of operational realities that compound over time. The first is hardware dependency. Your NPS server is a physical or virtual machine that requires patching, capacity planning, and eventual hardware refresh. In a multi-site deployment — say, a hotel group with properties across the UK — you're either running a centralised NPS with WAN dependency, or you're deploying NPS instances at each site and managing them individually. Neither is elegant. The second is availability. A single NPS instance is a single point of failure for your entire authentication infrastructure. Yes, you can deploy NPS in a failover pair, but that doubles your hardware and licensing overhead, and it still doesn't give you the geographic redundancy that a cloud service provides natively. The third is scalability. NPS was designed for corporate LAN environments. When you're handling thousands of concurrent authentication requests during a stadium event or a conference centre peak, the throughput limitations of a single NPS instance become very apparent. Authentication latency spikes, and users experience connection failures at exactly the moment you can least afford it. RADIUS as a Service addresses all three of these constraints architecturally. The cloud RADIUS provider runs a distributed, geo-redundant cluster of RADIUS servers. Your access points point to cloud-hosted RADIUS endpoints rather than an on-premises server. Authentication requests are load-balanced across the cluster, and failover is automatic and transparent. The provider handles patching, capacity scaling, and certificate management. From your perspective as the network operator, RADIUS becomes a consumed service rather than a managed component. The authentication protocols themselves don't change. You're still running IEEE 802.1X with EAP-TLS, PEAP-MSCHAPv2, or EAP-TTLS depending on your client device mix. The difference is where the RADIUS server lives and who is responsible for its operational continuity. There's an important security consideration here that I want to address directly, because it comes up in almost every client conversation. Moving RADIUS to the cloud means your authentication traffic is traversing the public internet to reach the cloud RADIUS endpoint. This is mitigated through two mechanisms. First, RADIUS traffic between the Network Access Server and the RADIUS server is protected using a shared secret and MD5-based message authentication. Second, and more importantly for modern deployments, you should be running RadSec — RADIUS over TLS, defined in RFC 6614 — which wraps the entire RADIUS conversation in a TLS tunnel. This gives you transport-layer encryption equivalent to HTTPS, eliminating the MD5 vulnerability and providing mutual authentication between the NAS and the RADIUS server. Any cloud RADIUS provider worth considering should support RadSec as standard. On the identity integration side, cloud RADIUS services typically support LDAP and LDAPS connections back to your on-premises Active Directory, or native integration with Azure Active Directory and Entra ID via SAML or SCIM. This means you don't need to migrate your user directory — the cloud RADIUS service queries your existing identity store, maintaining your existing user lifecycle management processes. For compliance-conscious organisations — and that includes anyone handling payment card data under PCI DSS, or personal data under GDPR — cloud RADIUS providers that are SOC 2 Type II certified and ISO 27001 accredited provide a stronger compliance posture than most organisations can achieve with self-managed NPS infrastructure. --- SEGMENT 3: IMPLEMENTATION RECOMMENDATIONS AND PITFALLS Right, let's talk about how you actually execute this migration without taking your authentication infrastructure offline. The methodology I recommend is a five-phase approach. Phase one is audit and inventory. Document every RADIUS client — every access point, every switch, every VPN concentrator — along with its current shared secret, the EAP method it's using, and any vendor-specific attributes in your NPS policies. This is the unglamorous work, but skipping it is the number one cause of migration failures. Phase two is pilot deployment. Stand up your cloud RADIUS instance and point a non-production SSID or a single test site at it. Validate that your EAP method works end-to-end, that your identity integration is functioning, and that your accounting data is flowing correctly. Phase three is parallel running. This is the critical risk mitigation step. Configure your access points with both the on-premises NPS server and the cloud RADIUS server as authentication targets, with the cloud service as primary and NPS as fallback. Run in this configuration for a minimum of two weeks across a full business cycle. Monitor authentication success rates, latency, and any policy discrepancies. Phase four is cutover. Remove the NPS fallback configuration and commit to cloud RADIUS as your sole authentication infrastructure. Do this during a planned maintenance window, and have a rollback procedure documented and tested. Phase five is decommission. Once you've validated stable operation for thirty days post-cutover, decommission the NPS servers and reclaim the hardware or virtual machine resources. The pitfalls I see most frequently are: certificate trust chain issues — specifically, client devices that don't trust the cloud RADIUS server's certificate because the CA isn't in their trusted store. Resolve this through your MDM or Group Policy before cutover. The second common pitfall is firewall rules. Cloud RADIUS requires outbound UDP 1812 and 1813 from your access points to the cloud endpoints, or TCP 2083 for RadSec. Ensure your network perimeter allows this traffic. Third: shared secret complexity. If your existing NPS shared secrets are weak, use the migration as an opportunity to rotate to cryptographically strong secrets, or better yet, move to RadSec and eliminate shared secrets entirely. --- SEGMENT 4: RAPID-FIRE Q&A Let me run through the questions I get most often on this topic. Can we keep Active Directory on-premises? Yes, absolutely. Cloud RADIUS connects to your on-premises AD via LDAPS. Your directory stays where it is. What happens if our internet connection goes down? This is the key dependency shift. With cloud RADIUS, internet connectivity becomes a dependency for authentication. Mitigate this with redundant WAN links or a local RADIUS proxy that caches authentication for known devices during outages. Does this affect our PCI DSS compliance? Moving to a certified cloud RADIUS provider typically improves your compliance posture. Ensure your provider can supply SOC 2 Type II reports and is included in your annual QSA assessment scope. How long does a full migration take? For a single site, two to four weeks. For a multi-site estate of fifty or more locations, plan for three to six months with a phased rollout. --- SEGMENT 5: SUMMARY AND NEXT STEPS To wrap up: the case for migrating from on-premises NPS to RADIUS as a Service is compelling on operational, financial, and compliance grounds. The migration itself is low-risk when executed with a structured parallel-running phase. The key technical decisions are your EAP method selection, your identity integration approach, and whether to implement RadSec for transport security — which I'd strongly recommend for any new deployment. Your immediate next steps: conduct the audit of your current RADIUS clients and policies, engage your cloud RADIUS provider for a pilot environment, and review your firewall rules and certificate trust chains before you start. For organisations running Purple WiFi's guest access platform, the RADIUS as a Service capability integrates directly with the guest WiFi authentication flow, giving you a single control plane for both corporate 802.1X authentication and guest network access management — with the analytics and compliance reporting built in. Thanks for listening. The full technical reference guide is available on the Purple website, and our solutions team is available for a scoping conversation if you're ready to move forward. --- END OF SCRIPT

header_image.png

Résumé Exécutif

Pendant près de deux décennies, le Network Policy Server (NPS) de Microsoft a été l'implémentation RADIUS par défaut pour les réseaux d'entreprise. Cependant, à mesure que les opérateurs de sites se développent sur des emplacements distribués – des chaînes de magasins aux groupes hôteliers mondiaux – le fardeau opérationnel de la gestion de l'infrastructure d'authentification sur site est devenu un passif important.

La migration vers RADIUS as a Service transforme l'authentification d'un composant matériel géré en un service cloud consommé. Cette transition architecturale élimine le point de défaillance unique inhérent aux déploiements NPS autonomes, supprime les cycles de renouvellement matériel et offre l'évolutivité élastique requise pour les environnements à haute densité comme les stades et les centres de conférence. Pour les responsables informatiques et les architectes réseau, ce guide fournit une méthodologie structurée et neutre vis-à-vis des fournisseurs pour migrer l'authentification 802.1X vers le cloud sans impacter le trafic de production, garantissant la conformité avec PCI DSS et GDPR, et réduisant les dépenses d'exploitation de l'infrastructure d'authentification jusqu'à 80 %.

Approfondissement Technique : Architecture et Normes

Pour comprendre la migration, nous devons d'abord examiner le changement architectural dans la manière dont le contrôle d'accès basé sur les ports IEEE 802.1X est fourni.

Les Limites du NPS sur Site

Dans un déploiement traditionnel, les points d'accès agissent comme le Network Access Server (NAS), transmettant les requêtes d'authentification à un serveur NPS sur site. Le serveur NPS évalue les politiques de requête de connexion, valide les identifiants par rapport à un magasin d'identités (généralement Active Directory via LDAP) et renvoie un message Access-Accept ou Access-Reject.

Ce modèle présente trois contraintes critiques pour les réseaux modernes :

  1. Dépendance et Maintenance Matérielles : NPS nécessite des machines physiques ou virtuelles dédiées, exigeant des correctifs continus, une planification de capacité et une gestion du cycle de vie.
  2. Complexité de la Haute Disponibilité : Atteindre la redondance nécessite le déploiement de NPS en paire de basculement, doublant les coûts de licence sans fournir une véritable redondance géographique.
  3. Goulots d'Étranglement de Débit : Pendant les pics de concurrence – comme l'entrée d'un stade ou les heures de pointe commerciales – une seule instance NPS peut devenir un goulot d'étranglement, entraînant des délais d'authentification et une expérience utilisateur dégradée.

L'Architecture RADIUS Cloud

RADIUS as a Service abstrait la couche d'authentification. Les fournisseurs de cloud exploitent des clusters de serveurs RADIUS distribués et géo-redondants. Le NAS pointe vers ces points de terminaison cloud, et les requêtes sont automatiquement équilibrées en charge.

architecture_comparison.png

Sécurité du Transport : Le Rôle de RadSec Lors du déplacement de RADIUS vers le cloud, le trafic d'authentification traverse l'internet public. Alors que RADIUS traditionnel utilise un secret partagé et le hachage MD5, les déploiements modernes doivent implémenter RadSec (RADIUS over TLS, RFC 6614). RadSec encapsule l'intégralité de la conversation RADIUS dans un tunnel TLS (généralement le port TCP 2083), fournissant un chiffrement de couche de transport équivalent à HTTPS et une authentification mutuelle entre le NAS et le point de terminaison RADIUS cloud.

Intégration d'Identité RADIUS cloud ne nécessite pas la migration de votre annuaire d'utilisateurs. Les services prennent généralement en charge les connexions LDAPS vers Active Directory sur site ou les intégrations API natives avec Azure Active Directory (Entra ID) via SAML ou SCIM. Cela garantit que vos processus de gestion du cycle de vie des utilisateurs existants restent intacts.

Pour les sites utilisant des plateformes Guest WiFi , RADIUS cloud s'intègre directement, offrant un plan de contrôle unifié pour l'authentification d'entreprise 802.1X et l'accès au réseau invité, avec des WiFi Analytics avancés.

Guide d'Implémentation : Une Méthodologie en 5 Phases

L'exécution d'une migration sans interruption nécessite une approche structurée et progressive.

migration_checklist.png

Phase 1 : Audit et Inventaire

Avant d'apporter des modifications, documentez l'état actuel :

  • Clients RADIUS : Identifiez chaque NAS (points d'accès, commutateurs, concentrateurs VPN).
  • Politiques : Documentez les politiques de requête de connexion et de réseau NPS existantes, y compris les attributs spécifiques au fournisseur (VSAs) utilisés pour l'affectation de VLAN.
  • Méthodes EAP : Identifiez les méthodes Extensible Authentication Protocol en cours d'utilisation (par exemple, EAP-TLS, PEAP-MSCHAPv2).

Phase 2 : Déploiement Pilote

Provisionnez l'instance RADIUS cloud et configurez un SSID non-production ou un seul site de test. Validez l'intégration de l'annuaire d'identités (par exemple, la synchronisation Entra ID) et assurez-vous que la méthode EAP fonctionne de bout en bout.

Phase 3 : Fonctionnement en Parallèle (Atténuation des Risques)

Configurez les périphériques NAS de production pour utiliser à la fois le serveur RADIUS cloud (Principal) et le serveur NPS hérité (Secours). Exécutez cette configuration pendant au moins deux semaines. Surveillez les taux de réussite d'authentification, les métriques de latence et les flux de données de comptabilité pour identifier toute divergence de politique avant le basculement.

Phase 4 : Basculement

Lors d'une fenêtre de maintenance planifiée, supprimez la configuration de secours NPS héritée des périphériques NAS. Engagez-vous entièrement dans l'infrastructure cloud. Assurez-vous que votre procédure de restauration est documentée et testée.

Phase 5 : Décommissionnement

Après 30 jours de fonctionnement stable, décommissionnez en toute sécurité les serveurs NPS hérités et récupérez les ressources de calcul.

Bonnes Pratiques et Conformité

Lors de la conception de votre architecture RADIUS cloud, respectez les normes suivantes :

  • Exiger RadSec : N'envoyez jamais de trafic RADIUS sur l'internet public en utilisant les ports UDP standard 1812/1813 si RadSec (TCP 2083) est pris en charge par votre matériel NAS.
  • Chaînes de Confiance des Certificats : Assurez-vous que les périphériques clients font confiance au CertAutorité de certification (CA) qui a émis le certificat du serveur RADIUS cloud. Poussez l'autorité de certification racine vers les appareils gérés via MDM ou une stratégie de groupe avant la migration.
  • Posture de conformité : Sélectionnez un fournisseur RADIUS cloud qui maintient la certification SOC 2 Type II et l'accréditation ISO 27001. Cela simplifie considérablement vos évaluations annuelles PCI DSS, en particulier pour les environnements de Commerce de détail et d' Hôtellerie .

Pour des principes de conception de réseau plus larges, consultez nos guides sur La configuration du WiFi pour les entreprises : Un guide stratégique pour 2026 et Comprendre le RSSI et la force du signal pour une planification optimale des canaux .

Dépannage et atténuation des risques

Mode de défaillance Cause première Stratégie d'atténuation
Délais d'authentification Pare-feu bloquant le trafic sortant UDP 1812/1813 ou TCP 2083. Vérifiez que les règles du pare-feu périmétrique autorisent le trafic sortant vers les plages d'adresses IP spécifiques du fournisseur RADIUS cloud.
Erreurs de confiance de certificat Les appareils clients ne disposent pas de l'autorité de certification racine dans leur magasin de confiance. Déployez l'autorité de certification racine via MDM/GPO avant la phase 3 (Fonctionnement en parallèle).
Échec d'attribution de VLAN Attributs spécifiques au fournisseur (VSA) non mappés correctement dans les politiques cloud. Reproduisez les formats de chaîne VSA exacts de NPS vers le moteur de politique RADIUS cloud pendant la phase 1.
Impact d'une panne WAN La perte d'internet entraîne la perte d'accès au RADIUS cloud. Déployez des liens WAN redondants ou implémentez un proxy RADIUS local qui met en cache les identifiants des appareils connus.

Retour sur investissement et impact commercial

La migration vers RADIUS as a Service offre des résultats commerciaux mesurables :

  • Réduction des coûts : Élimine l'acquisition de matériel, les licences Windows Server et les heures d'ingénierie consacrées aux correctifs et à la maintenance. La réduction typique des OpEx est de 60 à 80 %.
  • SLA de fiabilité : Les fournisseurs cloud offrent des SLA de disponibilité de 99,99 % garantis financièrement, comparés aux 97-98 % typiques atteints par les déploiements NPS sur site unique.
  • Agilité : De nouveaux sites peuvent être mis en ligne instantanément sans provisionnement de matériel d'authentification local, accélérant les délais de déploiement pour les pôles de Transport et les établissements de Santé .

Écoutez notre équipe de consultants seniors discuter des implications stratégiques dans ce briefing de 10 minutes :

Définitions clés

RADIUS (Remote Authentication Dial-In User Service)

A networking protocol that provides centralized Authentication, Authorization, and Accounting (AAA) management for users who connect and use a network service.

The core protocol used by enterprise WiFi networks to validate user credentials before granting network access.

NPS (Network Policy Server)

Microsoft's implementation of a RADIUS server and proxy, bundled as a role in Windows Server.

The legacy on-premises infrastructure that organizations are actively migrating away from to reduce maintenance overhead.

NAS (Network Access Server)

The device that acts as the gateway to the network and passes authentication requests to the RADIUS server.

In a wireless context, the NAS is typically the WiFi Access Point or Wireless LAN Controller.

RadSec (RADIUS over TLS)

A protocol defined in RFC 6614 that transports RADIUS packets over a TCP connection encrypted with TLS.

Essential for cloud RADIUS deployments to ensure credential data is encrypted while traversing the public internet.

EAP (Extensible Authentication Protocol)

An authentication framework frequently used in wireless networks and point-to-point connections.

Determines how the client and server securely exchange credentials (e.g., certificates via EAP-TLS, or passwords via PEAP).

VSA (Vendor-Specific Attribute)

Custom attributes defined by hardware vendors within the RADIUS protocol to support proprietary features.

Crucial during migration; VSAs are often used to assign authenticated users to specific network VLANs dynamically.

LDAPS (Lightweight Directory Access Protocol over SSL)

A secure protocol for querying and modifying directory services like Active Directory.

Used by cloud RADIUS services to securely query on-premises identity stores without migrating the user directory to the cloud.

802.1X

An IEEE standard for port-based network access control (PNAC).

The underlying standard that uses RADIUS to ensure only authenticated devices can pass traffic onto the enterprise LAN or WLAN.

Exemples concrets

A 200-property hotel group currently runs local NPS servers at each site for staff 802.1X authentication. They are migrating to Entra ID (Azure AD) and want to decommission the local servers. How should they approach the migration?

  1. Deploy a cloud RADIUS service that integrates natively with Entra ID via SAML/SCIM.
  2. Configure the cloud RADIUS policies to map Entra ID groups (e.g., 'Front Desk', 'Management') to specific VLAN VSAs.
  3. At a pilot property, configure the access points to use RadSec to connect to the cloud RADIUS endpoint.
  4. Push the cloud RADIUS server's Root CA to all staff devices via Microsoft Intune.
  5. Run parallel authentication at the pilot site, then execute a phased rollout across the remaining 199 properties.
Commentaire de l'examinateur : This approach removes 200 physical/virtual servers from the estate, drastically reducing the attack surface and maintenance overhead. Integrating directly with Entra ID eliminates the need for complex site-to-site VPNs back to a central Active Directory.

A stadium with 50,000 capacity experiences authentication failures on their corporate SSID during major events because their on-premises NPS server cannot handle the throughput of thousands of devices roaming simultaneously.

  1. Audit the existing NPS policies and EAP methods.
  2. Provision a cloud RADIUS service capable of auto-scaling to handle high authentications per second (APS).
  3. Establish an LDAPS connection from the cloud RADIUS service to the stadium's on-premises Active Directory.
  4. Update the stadium's high-density wireless LAN controllers to point to the cloud RADIUS endpoints as the primary authentication servers.
Commentaire de l'examinateur : By offloading the RADIUS processing to a cloud cluster, the stadium leverages elastic compute resources that scale dynamically during event ingress, resolving the bottleneck without requiring the venue to over-provision expensive local hardware.

Questions d'entraînement

Q1. Your organization is migrating to Cloud RADIUS. The security team mandates that no authentication traffic can be sent over the internet in cleartext or using deprecated hashing algorithms like MD5. What protocol must you configure on your wireless LAN controllers?

Conseil : Look for the protocol that wraps RADIUS in a TLS tunnel.

Voir la réponse type

You must configure RadSec (RADIUS over TLS). RadSec establishes a TLS tunnel over TCP port 2083 between the NAS and the cloud RADIUS server, providing transport-layer encryption and mutual authentication, satisfying the security team's requirements.

Q2. During Phase 3 (Parallel Running) of your migration, you notice that users are authenticating successfully against the cloud RADIUS server, but they are not being placed in the correct network segments. What is the most likely configuration gap?

Conseil : How does a RADIUS server tell an access point which network segment to use?

Voir la réponse type

The Vendor-Specific Attributes (VSAs) for dynamic VLAN assignment have not been configured correctly in the cloud RADIUS policies. You must ensure the exact VSA strings used in the legacy NPS server are replicated in the cloud environment so the NAS knows which VLAN to assign to the user.

Q3. A client device is repeatedly failing EAP-TLS authentication against the new cloud RADIUS service, but it works fine against the legacy NPS server. The device logs show an 'untrusted server' error. How do you resolve this?

Conseil : EAP-TLS requires the client to trust the server's identity.

Voir la réponse type

The client device does not have the Root Certificate Authority (CA) that issued the cloud RADIUS server's certificate in its trusted root store. You must deploy the Root CA to the client device using a Mobile Device Management (MDM) solution or Group Policy.