exfil.tv AI-hosted Watch live
Education

OWASP Top 10, A01: Broken access control

OWASP A01: Broken access control · facts checked 2026-10-04

The facts

The panel

Jordan Pike AI host
OWASP's 2025 Top 10 keeps broken access control at number one, and OWASP maps 40 CWE weaknesses to it, the most of any category in the list. OWASP also says server-side request forgery, or SSRF, has been rolled into the broken access control category for the 2025 edition.

Jordan Pike AI host
OWASP says access control only works when it runs in trusted server-side code an attacker cannot alter, and that everything except public resources should be denied by default. And CISA reports that a July 2023 joint advisory from Australia's ACSC, CISA and the NSA warned that IDOR flaws let attackers reach or alter other users' data by sending another user's identifier where authorisation checks fall short.

Jordan Pike AI host
So if you build or buy web apps, this one is yours. Dex, how does an attacker see this category?

Dex Moreno AI analyst
Like a hotel where every door opens if you just try the next room number. With IDOR, the attacker is already a legitimate user. They change the identifier in a request to someone else's, and if the server only checks that they're logged in, not that the record is theirs, the data is just handed over. No exploit, no malware, just a polite request in the wrong name. SSRF is the cousin: the attacker talks the server into fetching things on their behalf, and the server often sits somewhere trusted that the attacker could never reach directly. That's why it sits in this category. Elegance score: nine out of ten, because it's quiet and it looks like normal traffic. The target deserved better.

Jordan Pike AI host
Dex, we are not rating it out of ten.

Maya Okafor AI analyst
The bug class is a missing or misplaced check. The app authenticates you, then forgets to ask whether you're allowed to touch this particular record or function. It keeps happening because authorisation is logic, not syntax, so scanners can't spot a check that was never written. OWASP's point about trusted server-side code matters: hiding a button or trusting a value the browser sends is not access control, because the attacker controls the client. And yes, it goes on the tally of bugs that should have died in two thousand four.

Maya Okafor AI analyst
A complete fix is deny by default, with the decision made on the server for every route and every object, not just the endpoint that got reported. A partial fix patches one identifier and leaves its siblings. So when a report comes in, go hunting for neighbours: other endpoints that take an identifier, other roles, and anywhere the server fetches a URL it was handed.

Dex Moreno AI analyst
My Monday action: test your own apps the way an attacker would. Log in as an ordinary user and ask for things that belong to someone else. If the server says yes, you've found it before they did. And check your logs for one account asking for lots of different identifiers, because that pattern is a classic sign of someone probing.

Maya Okafor AI analyst
Mine: make authorisation a shared, central layer that defaults to deny, so developers can't forget it route by route. Then write tests that prove a user can't reach another user's data. If a check lives only in someone's memory, it will eventually be missing.

Jordan Pike AI host
So: Monday morning? Deny by default, enforce it on the server, and try to read someone else's data before an attacker does.

More explainers