The attack doesn't rely on redirecting to unregistered URLs, that's the problem. The goal of the attack is to circumvent phishing filters, by presenting a URL from a legitimate domain (the AS) that eventually redirects to the actual phishing URL. The actual phishing page doesn't need to target the same authorization server, or an authorization server at all for that matter. An attacker can register a legitimate app on any authorization server as a service, on their own tenant. The goal is just to have a starting URL that phishing filters won't block, and the attacker is in full control of the redirect URIs they register in their own tenant.
My take: it might be tricky to change the redirect on error behavior at this point, but we should at least note the issue in the security considerations/BCPs and possibly give some advice. For example, on top of my head: AS should expose their endpoints on a domain dedicated to OAuth/OIDC operations, and avoid using its top level domains (different area/service, but think herokuapp.com vs heroku.com) so that if a phishing filter decides to block direct links to the issuing endpoints will only impact things like IdP initiated flows (solvable by adding jumpstart endpoints on the RP anyway, just like IdP initiated sign in works in OIDC). I am sure there are lots of other things we can come up with that can make the problem better. On Fri, Dec 17, 2021 at 5:00 AM Warren Parad <wparad= 40rhosys...@dmarc.ietf.org> wrote: > I think this just falls into the category of never redirect the user to a > url that doesn't match one of the preregistered redirect urls (or logout > urls for that matter). Any application that has redirects anywhere provides > an opportunity for this attack vector, OAuth isn't unique in that way, it > just is consistent and documented. And the 2.1 draft is pretty clear on > this front: > > > https://datatracker.ietf.org/doc/html/draft-ietf-oauth-v2-1-04#section-4.1.2.1 > >> If the request fails due to a missing, invalid, or mismatching >> redirect URI, or if the client identifier is missing or invalid, the >> authorization server SHOULD inform the resource owner of the error >> and >> *MUST NOT automatically redirect the user agent to the invalid redirect >> URI*. > > > I want to call this attack vector "*illegitimate* phishing applications" > which is easily blocked by preregistration and/or PARs. And is only a very > small subset of phishing attacks with OAuth, of which the larger group is " > *legitimate* phishing applications". An app can be registered correctly, > and still issue a phishing attack as phishing attacks through OAuth are > actually indistinguishable from standard user delegation. There is no way > to prevent these without an application review before registration is > completed, here's an example that cloned Google apps y creating a fake app > called *google defender*: > https://www.trendmicro.com/en_us/research/17/d/pawn-storm-abuses-open-authentication-advanced-social-engineering-attacks.html > > If we can't protect against these latter ones, I hardly think protecting > against the former is useful/interesting/valuable. > > Warren Parad > > Founder, CTO > Secure your user data with IAM authorization as a service. Implement > Authress <https://authress.io/>. > > > On Thu, Dec 16, 2021 at 9:05 PM Rifaat Shekh-Yusef < > rifaat.s.i...@gmail.com> wrote: > >> All, >> >> >> An article was recently published discussing some OAuth Redirection >> Attacks to try to bypass phishing detection solutions. See the details >> of these attacks in the following link: >> >> >> >> https://www.proofpoint.com/us/blog/cloud-security/microsoft-and-github-oauth-implementation-vulnerabilities-lead-redirection >> >> >> The article discusses attacks on Microsoft and GitHub, but these attacks >> are not unique to these companies. >> >> The attacks take advantage of how OAuth handles error responses, which >> sends responses to the application’s redirect URL. >> >> I would like to get the thoughts of the working group on these types of >> attacks. >> >> What is the best way to mitigate these attacks? >> >> Do we need a new approach for handling errors with OAuth? >> >> Regards, >> >> Rifaat >> >> _______________________________________________ >> OAuth mailing list >> OAuth@ietf.org >> https://www.ietf.org/mailman/listinfo/oauth >> > _______________________________________________ > OAuth mailing list > OAuth@ietf.org > https://www.ietf.org/mailman/listinfo/oauth >
_______________________________________________ OAuth mailing list OAuth@ietf.org https://www.ietf.org/mailman/listinfo/oauth
