Overview
A redirect endpoint may accept a next, return or url parameter and send the browser elsewhere. If the destination is not validated, attackers can use a trusted domain to make phishing links look legitimate.
Prefer internal route identifiers
The safest design is to accept a known route key rather than a complete URL. Map that key to a server-defined path. If a full destination is unavoidable, parse it and enforce a strict allow-list.
Reject external destinations by default
Do not rely on checks such as str_contains($url, 'example.com'). Inputs like example.com.attacker.test, user-info syntax and unusual URL forms can bypass naive string checks.
$allowedPaths = [
'dashboard' => '/account/dashboard',
'settings' => '/account/settings',
];
$key = $_GET['return'] ?? 'dashboard';
$destination = $allowedPaths[$key] ?? '/';
header('Location: ' . $destination, true, 303);
exit;
This approach avoids parsing an attacker-controlled absolute URL entirely.
If external destinations are required
Compare the parsed scheme, host and port against a fixed allow-list. Require HTTPS, reject credentials and unexpected ports, and normalize hostnames carefully. Consider internationalized domain names and trailing-dot behavior. A redirect allow-list should be small and maintained centrally.
Preserve safe navigation
For post-login redirects, store a validated local path in the session before authentication, then revalidate it before use. Never assume that a value is safe because it was previously stored in a cookie or hidden form field.
Checklist
- Prefer route keys mapped to internal paths.
- Reject protocol-relative URLs and unexpected schemes.
- Do not use substring checks for host validation.
- Revalidate redirect targets at the point of use.
- Use a central redirect helper.
- Test encoded, malformed and deceptive URL forms.