訪客 WiFi 的 Walled Garden 設定

PODCAST SCRIPT: Walled Garden Configuration for Guest WiFi Purple WiFi Intelligence Platform — Technical Briefing Series Duration: Approximately 10 minutes Voice: UK English, senior consultant tone — confident, conversational, authoritative --- [INTRO — 1 MINUTE] Welcome to the Purple Technical Briefing Series. I'm your host, and today we're tackling one of the most consistently misunderstood elements of enterprise guest WiFi deployment: the walled garden. If you've ever had a guest WiFi rollout where users couldn't get past the splash page — couldn't log in with Google, couldn't load the captive portal at all — there's a very good chance the walled garden configuration was the culprit. And yet, it's one of those things that rarely gets the attention it deserves in a network design brief. In the next ten minutes, I want to give you a clear, practical picture of what a walled garden actually is, why it matters, which domains you need to whitelist, and how social login integrations change the equation. Whether you're deploying guest WiFi across a hotel estate, a retail chain, a stadium, or a public sector estate, this briefing will give you the configuration framework you need to get it right first time. Let's get into it. --- [TECHNICAL DEEP-DIVE — 5 MINUTES] So, what is a walled garden in the context of guest WiFi? Think of it this way. When a guest device connects to your WiFi network but hasn't yet authenticated through your captive portal, that device is in a kind of limbo state. It has an IP address. It can send and receive packets. But your network controller — whether that's a Cisco Meraki, a Ruckus SmartZone, a Fortinet FortiGate, or a cloud-managed Aruba platform — is intercepting all outbound HTTP and HTTPS requests and redirecting them to your splash page. The walled garden is the set of domains and IP addresses that are explicitly permitted to pass through that interception layer before authentication completes. Everything else is blocked. That's the wall. The garden is the curated space inside it — the small, controlled set of resources a guest can reach before they've proven who they are. Now, why does this matter so much? Because modern captive portals are not self-contained. They depend on external services to function. Your splash page might be hosted on a CDN. Your social login buttons call OAuth endpoints at Google, Facebook, Apple, or Microsoft. Your analytics platform might be loading tracking scripts. Your payment gateway — if you're charging for premium access — needs to load its own JavaScript and make API calls. Every single one of those external dependencies needs to be explicitly whitelisted in your walled garden, or the authentication flow will break. Let me walk you through the categories of domains you need to consider. First: your captive portal platform itself. If you're using Purple, that means whitelisting the Purple CDN and API endpoints — things like cdn.purple.ai, portal.purple.ai, and api.purple.ai. If you're using a different platform, the principle is identical: whitelist every domain that serves the portal assets and handles the authentication handshake. Second: Google OAuth. This is the big one, because Google Sign-In is the most common social login method in enterprise guest WiFi deployments. You need to whitelist accounts.google.com, oauth2.googleapis.com, apis.google.com, and the gstatic.com CDN — that's where Google serves its fonts, icons, and client-side libraries. Miss any one of these and the Google login button will either fail silently or throw a CORS error that the guest never sees. Third: Facebook and Meta OAuth. If you're offering Facebook login — and in hospitality and retail, it remains popular because of the marketing data it provides — you need to whitelist www.facebook.com, graph.facebook.com, connect.facebook.net, and the Facebook CDN at static.xx.fbcdn.net. Meta has a habit of rotating CDN subdomains, so I'd recommend using wildcard entries where your controller supports them: *.fbcdn.net and *.facebook.com. Fourth: Apple Sign In. This became mandatory for any iOS application offering third-party login after 2019, and it's increasingly expected on web-based portals too. The key domains are appleid.apple.com and idmsa.apple.com. Apple also uses a range of subdomains under apple.com for its authentication flows, so a wildcard entry for *.apple.com is the pragmatic approach. Fifth: if you're running a paid WiFi tier — common in transport hubs, premium hotel properties, and conference centres — you need to whitelist your payment gateway. For Stripe, that's stripe.com and *.stripe.com. For PayPal, it's www.paypal.com and *.paypal.com. PCI DSS compliance requires that payment flows are handled over TLS 1.2 or higher, and your walled garden configuration needs to permit that traffic without interception. Now, here's where it gets technically interesting: the DNS resolution problem. Most network controllers implement walled gardens at the IP address level, not purely at the domain name level. That means when you whitelist accounts.google.com, the controller resolves that domain to a set of IP addresses and permits traffic to those IPs. The problem is that Google, Facebook, Apple, and the major CDNs use dynamic IP ranges and anycast routing. The IP address that accounts.google.com resolves to in your data centre is not necessarily the same IP it resolves to on your guest network, and it will change over time. The practical implication is this: you cannot rely on a static IP whitelist. You need a controller that performs dynamic DNS resolution for walled garden entries — resolving the whitelisted domains at regular intervals and updating the permitted IP set accordingly. Most enterprise-grade controllers support this. If yours doesn't, you're operating with a configuration that will degrade over time as CDN IP ranges shift. There's also the HTTPS interception question. When a guest device makes an HTTPS request to a non-whitelisted domain before authentication, your controller has two options: it can drop the connection silently, or it can attempt to intercept it and redirect to the portal. Silent drops cause the guest's browser to display a generic "site can't be reached" error, which is confusing. Interception requires a valid TLS certificate on your controller, and without it, the guest sees a certificate warning — which is both alarming and, in regulated environments, a potential compliance issue. The clean solution is to ensure your portal redirect logic operates on HTTP traffic, and that your walled garden permits the HTTPS traffic for whitelisted domains to pass through untouched. Let me also address the captive portal detection mechanism, because it directly affects your walled garden design. Modern operating systems — iOS, Android, macOS, Windows — use a technique called Captive Network Assistant, or CNA. When a device connects to a new network, the OS makes an HTTP request to a known probe endpoint: on Apple devices, that's captive.apple.com; on Android, it's connectivitycheck.gstatic.com; on Windows, it's msftconnecttest.com. If the response is not what the OS expects, it knows it's behind a captive portal and launches the portal browser automatically. The critical point: all of these probe endpoints must be whitelisted in your walled garden. If they're blocked, the OS never detects the captive portal, the guest never sees the splash page, and from their perspective the WiFi simply doesn't work. This is one of the most common misconfiguration failures I see in the field, and it's entirely avoidable. --- [IMPLEMENTATION RECOMMENDATIONS AND PITFALLS — 2 MINUTES] Let me give you the implementation framework I'd recommend for any new deployment. Start with a baseline whitelist that covers five categories: your portal platform, Google OAuth, Facebook OAuth, Apple Sign In, and OS captive portal probes. That's your minimum viable walled garden. Add payment gateways if you're running paid tiers. Add your analytics and marketing platform domains if your portal loads tracking scripts. Test your walled garden before go-live using a device in an unauthenticated state — not a test account, an actual fresh device that has never connected to this network. Walk through every login method you're offering. If any login method fails, capture the browser console output and network traffic to identify which domain is being blocked. Implement a quarterly review process. OAuth providers and CDNs change their domain structures. Apple updated its Sign In domains twice in 2023. Google periodically adds new subdomains to its OAuth flow. A walled garden that was correct at deployment will drift out of alignment without active maintenance. The pitfalls to avoid: first, over-whitelisting. I've seen deployments where the IT team, frustrated by intermittent failures, simply whitelisted entire IP ranges or added wildcard entries that effectively bypassed the walled garden entirely. That defeats the purpose and creates a compliance risk. Be precise. Second, ignoring IPv6. If your network supports IPv6 — and increasingly it should — your walled garden rules need to cover IPv6 address ranges as well as IPv4. Third, not accounting for mobile app deep links. Some social login flows, particularly on iOS, attempt to open the native app rather than a web browser. This can break the OAuth flow entirely. Ensure your portal configuration forces web-based OAuth rather than app-based flows. --- [RAPID-FIRE Q&A — 1 MINUTE] Let me run through a few questions I hear regularly. "Do I need to whitelist the entire Google IP range?" No. Whitelist specific domains and use dynamic DNS resolution. Whitelisting entire ASNs is a security risk. "Can I use the same walled garden config across all my sites?" In principle, yes — if your portal platform and social login providers are consistent. But test at each site, because local DNS resolvers can behave differently. "How does GDPR affect walled garden configuration?" GDPR doesn't directly govern walled garden domains, but it does govern what data your portal collects during authentication. Ensure your social login OAuth scopes request only the minimum necessary data — typically name and email — and that your privacy notice is accessible from within the walled garden before the guest authenticates. "What's the right TTL for DNS entries in the walled garden?" Most controllers default to 60 seconds. For high-availability deployments, I'd recommend no lower than 30 seconds to avoid excessive DNS query load. --- [SUMMARY AND NEXT STEPS — 1 MINUTE] To summarise: a walled garden is the controlled pre-authentication zone in your guest WiFi deployment. Getting it right means whitelisting your portal platform, all social OAuth providers you're using, OS captive portal probe endpoints, and any payment or analytics services your portal depends on. Use dynamic DNS resolution, not static IP lists. Test with real unauthenticated devices before go-live. And build a quarterly review process into your operational calendar. If you're deploying or reviewing a guest WiFi estate and want to validate your current walled garden configuration, Purple's platform includes built-in walled garden management with pre-configured domain sets for all major social login providers. It's one of those things that's genuinely easier to get right with the right tooling behind you. Thanks for listening. The full technical reference guide for this topic — including architecture diagrams, domain whitelists, and worked implementation scenarios — is available in the Purple knowledge base. Until next time. --- [END OF SCRIPT]