Almost every incident worth writing about starts with something that was working exactly as designed. The design was just describing a different world than the one the system ended up in.
This piece walks through how we think about it at Paycux, what we have changed our minds about, and where the sharp edges are.
Validate every part of the assertion
That brings us to validate every part of the assertion. Detection beats prevention when prevention is uncertain. If you cannot stop a class of abuse outright, make sure it produces a signal loud enough that someone sees it the same day.
Blast radius is the metric that actually matters. A credential that can read one organization's data is an incident; the same credential scoped to every organization is a company-ending event, and the difference is a design decision made months earlier.
Use a well-maintained SAML library
Use a well-maintained SAML library deserves its own treatment. Detection beats prevention when prevention is uncertain. If you cannot stop a class of abuse outright, make sure it produces a signal loud enough that someone sees it the same day.
Detection beats prevention when prevention is uncertain. If you cannot stop a class of abuse outright, make sure it produces a signal loud enough that someone sees it the same day.
- Scope credentials to the smallest useful blast radius
- Alert on the abuse you cannot prevent
- Exercise the rotation path before you need it
- Log the decision, not just the outcome
Reject SHA-1 and other weak algorithms
Consider reject sha-1 and other weak algorithms. Detection beats prevention when prevention is uncertain. If you cannot stop a class of abuse outright, make sure it produces a signal loud enough that someone sees it the same day.
Blast radius is the metric that actually matters. A credential that can read one organization's data is an incident; the same credential scoped to every organization is a company-ending event, and the difference is a design decision made months earlier.
Implement replay attack prevention
Implement replay attack prevention deserves its own treatment. Blast radius is the metric that actually matters. A credential that can read one organization's data is an incident; the same credential scoped to every organization is a company-ending event, and the difference is a design decision made months earlier.
Rotate on a schedule, not on an incident. A rotation path that has been exercised is a routine change; one that has never been used is an outage waiting for the worst possible moment.
Blast radius is the metric that actually matters.
Manage certificates carefully
That brings us to manage certificates carefully. Blast radius is the metric that actually matters. A credential that can read one organization's data is an incident; the same credential scoped to every organization is a company-ending event, and the difference is a design decision made months earlier.
Blast radius is the metric that actually matters. A credential that can read one organization's data is an incident; the same credential scoped to every organization is a company-ending event, and the difference is a design decision made months earlier.
Be deliberate about IdP-initiated SSO
Be deliberate about IdP-initiated SSO deserves its own treatment. Detection beats prevention when prevention is uncertain. If you cannot stop a class of abuse outright, make sure it produces a signal loud enough that someone sees it the same day.
Rotate on a schedule, not on an incident. A rotation path that has been exercised is a routine change; one that has never been used is an outage waiting for the worst possible moment.
Scope group and role claims tightly
Scope group and role claims tightly is where this gets concrete. Rotate on a schedule, not on an incident. A rotation path that has been exercised is a routine change; one that has never been used is an outage waiting for the worst possible moment.
Detection beats prevention when prevention is uncertain. If you cannot stop a class of abuse outright, make sure it produces a signal loud enough that someone sees it the same day.
Where this leaves us
The pattern repeats across every system we have looked at: the hard part is not the mechanism, it is keeping the mechanism honest as the surrounding assumptions change.
If you are working through the same problem and want to compare notes, the docs cover the mechanics and the console shows the behaviour on your own data.
Everything here, already built
Sign-in, enterprise SSO, directory provisioning, roles and an audit trail behind one API. Start with the quickstart and have a working sign-in this afternoon.