squirrelscan
GitHub

Open redirects: when a redirect parameter sends visitors anywhere

A next, url or returnTo parameter that redirects to any host it is given lends your domain to phishing links. squirrel sends an off-site value that cannot resolve and reads the Location header without following it.

Why an open redirect is a problem

Sign-in pages, language switchers and tracking links often take the page to go to next from the URL: /login?next=/account. When the server redirects to whatever that parameter says, anyone can build a link that starts on your domain and ends on theirs:

https://example.com/login?next=https://evil.example/

The link shows your domain, passes a glance and many link filters, and lands the visitor on a look-alike sign-in page. In an OAuth or single sign-on flow the same flaw can hand a token or authorization code to the other site. The OWASP cheat sheet on unvalidated redirects covers the attack in detail.

What squirrel checks

This is a probing rule. It sends requests beyond the crawl, so it runs only when probing intensity is active or aggressive. On a passive run it sends nothing and records one skipped check.

Which URLs are probed

The rule looks for redirect-shaped query parameters in the URLs the audit already found: the crawled page URLs, the internal links on those pages, and the GET call sites and form actions that endpoint discovery found in your pages and scripts. Finding them sends no request. A parameter is redirect-shaped when its name is one of next, url, redirect, redirect_url, redirect_to, return, return_url, return_to, return_path, dest, destination, continue, goto or r. Names match without case, _ or -, so returnTo, return_to and return-to are the same.

Every URL must have the same origin as the audited URL. A path that names a state-changing action, such as logout, delete, update or reset, is never probed, and neither is a discovered call site or form that declares any method other than GET. Below aggressive, a path that robots.txt disallows is skipped. Each path and parameter name is probed once, and at most ten are probed per audit.

What is sent

For each URL and parameter, the same URL with only that parameter replaced:

GET /login?next=https://squirrelscan-redirect-probe.invalid/
GET /login?next=//squirrelscan-redirect-probe.invalid/

The second request, the protocol-relative form, is sent only when the first was not redirected off-site. It catches a check that refuses a scheme but lets //host through because it starts with /. The host is under .invalid, a name reserved never to resolve, so even a browser that followed the link would go nowhere.

Requests use GET only, carry no cookies and never follow the redirect: the rule reads the status and the Location header and does not read the body. They share the audit’s probe budget with the other probing rules.

Results

The rule emits checks named open-redirect:

  • Skipped when probing is passive. Message: Probing is passive, so no redirect parameter was probed. Run with --probe active to check for open redirects.
  • Skipped, with reason probe-budget, when the shared probe budget ran out before the first request, or partway through with no finding on the parameters it reached. A run cut short is never reported as a pass.
  • Skipped, with reason probe-errors, when every probe failed with a network error or timeout.
  • Warn, one check per page and parameter, when the answer is a 3xx whose Location resolves to the probe host. Message: Open redirect at https://example.com/login: the next parameter sends visitors to any site it names (a 302 to an off-site host given as an absolute URL). The finding names the page by origin and path only: the rest of the query string is never reported.
  • Pass when no parameter redirected off-site. Message: No redirect parameter sent visitors off-site (4 parameter(s) probed).

What is not flagged

  • A redirect that stays on your site, even when its query string repeats the value.
  • A redirect to any other host, such as an allowlisted partner or a fixed sign-in service. Only the probe host proves the value was used as given.
  • A 200 page that prints the value, a 400 or any other answer that is not a redirect.

Limits

The rule reads the Location header only. A page that redirects with a meta refresh or JavaScript after loading is not checked, and neither is a parameter the audit never saw in a URL, such as one that only a script adds.

How to fix it

Never redirect to a value from the request as it is. Accept only a relative path that starts with a single /, and reject //, /\ and anything with a scheme. Better still, map a short key to a fixed list of destinations, or parse the value with a URL parser and check its host against a list of your own hosts. A string check such as startsWith("/") is not enough, because //evil.example passes it. When the value fails the check, redirect to a safe default page.

Rule IDsecurity/open-redirect
CategorySecurity
ScopeSite-wide
Severitywarning
Weight5/10

Enable / disable

Disable this rule

[rules]
disable = ["security/open-redirect"]

Disable all Security rules

[rules]
disable = ["security/*"]

Enable only this rule

[rules]
enable = ["security/open-redirect"]

enable replaces the default ["*"], so this is the only rule selected. Leave disable out: a rule that matches any disable pattern is skipped even when enable names it. Only an enabled setting under a rule’s own [rule_options] table overrides both lists.

Check names by outcome

Each check name this rule can emit, by what it means to a tool comparing two audits. A problem check is a finding. A clean check says the rule ran over what it covers and found nothing; it clears the problem checks listed beside it, so a finding is fixed when the rule ran over the same scope and a clean check took its place. A not checked check says the rule did not evaluate, and never clears a finding.

Check nameOutcomeClears
open-redirectProblem
open-redirectCleanopen-redirect
open-redirectNot checked

Type to search…

↑↓ navigate↵ openesc close