Captive portal detection

Captive portals are used by public Wi-Fi networks (such as airports, coffee shops, and hotels) to make a user agree to their Terms of Service or provide payment before allowing access to the Internet. When a user connects to the Wi-Fi, the captive portal blocks all HTTPS traffic until the user completes a captive portal login flow in their browser. This prevents the WARP client from connecting to Cloudflare. At the same time, WARP creates firewall rules on the device to send all traffic to Cloudflare. The user is therefore unable to access the captive portal login screen unless they temporarily disable WARP.

Allow users to connect to captive portals

To allow users to connect through a captive portal, administrators can configure the following WARP settings:

No user interaction required

• Enable Captive portal detection. This allows WARP to temporarily turn off when it detects a captive portal on the network. For more details, refer to how captive portal detection works and its limitations.
• Set Device tunnel protocol to MASQUE. When using MASQUE, WARP traffic will look like standard HTTPS traffic and is therefore less likely to be blocked by captive portals.

User interaction required

• Enable Lock WARP switch and enable Admin override. Users can contact the IT administrator for a one-time code that allows them to manually turn off WARP and connect to a portal.
• For employees who travel, disable Lock WARP switch and set an Auto connect duration. This allows the user to manually turn off WARP without contacting IT.

How captive portal detection works

If WARP cannot establish a connection to Cloudflare, it will:

1. Temporarily open the system firewall so that the device can send traffic outside of the WARP tunnel. The firewall only allows the following traffic:

   • HTTP/HTTPS on TCP ports 80, 443, 8080, and 8443
   • DNS on UDP port 53

2. Send a series of requests to the captive portal test URLs. If the HTTPS request is intercepted, WARP assumes the network is behind a captive portal.

3. Open a browser window with the captive portal login screen if the captive portal sends a redirect HTTP response code (302, 303, 307, or 308).

4. Automatically re-enable the firewall after the configured timeout period.

flowchart TB
accTitle: Captive portal detection


A[Send DNS request] --Succeed--> B[Send HTTPS request]--Fail--> C[Send HTTP request] --Succeed--> D[Captive portal detected]--Receive HTTP redirect-->I[Redirect to captive portal login]
A --Fail--> F(CF_NO_NETWORK error)
C --Fail--> F
B --Succeed--> G[No captive portal]--> H[Retry connection to Cloudflare]

Limitations

• Due to how captive portal detection works, it may be possible for an employee to spoof a captive portal in order to turn off WARP.

• Some captive portals, particularly those on airlines, may be slow to respond and exceed the captive portal detection timeout. Users will likely see a CF_CAPTIVE_PORTAL_TIMED_OUT error when they try to connect.

• WARP may not be able to detect multi-stage captive portals, which redirect the user to different networks during the login process. Users will need to manually turn off WARP to get through the captive portal.

• Some public Wi-Fi networks are incompatible with running WARP:

  • Captive portals that intercept all DNS traffic will block WARP’s DoH connection. Users will likely see a CF_NO_NETWORK error after they login to the captive portal.
  • Captive portals that only allow HTTPS traffic will block WARP’s Wireguard UDP connection. Users will likely see a CF_HAPPY_EYEBALLS_MITM_FAILURE error after they login to the captive portal.