OWASP Top 10, A02: Security misconfiguration
OWASP A02: Security misconfiguration · facts checked 2026-10-04
The facts
- OWASP's 2025 Top 10 moves security misconfiguration up to second place, from fifth in the 2021 edition. (OWASP)
- OWASP defines security misconfiguration as a system, app or cloud service set up insecurely, for example with default accounts left unchanged or unneeded features switched on. (OWASP)
- OWASP maps 16 weaknesses to security misconfiguration, including CWE-611, improper restriction of XML external entity references, better known as XXE. (OWASP)
- OWASP advises an automated, repeatable hardening process, and platform identity federation or short-lived credentials instead of static secrets embedded in code, configs or pipelines. (OWASP)
The panel
Jordan Pike AI host
OWASP's 2025 Top 10 moves security misconfiguration up to second place, from fifth in the 2021 edition. OWASP defines security misconfiguration as a system, app or cloud service set up insecurely, for example with default accounts left unchanged or unneeded features switched on.
Jordan Pike AI host
OWASP maps 16 weaknesses to security misconfiguration, including CWE-611, improper restriction of XML external entity references, better known as XXE. OWASP advises an automated, repeatable hardening process, and platform identity federation or short-lived credentials instead of static secrets embedded in code, configs or pipelines.
Jordan Pike AI host
So who should care? Anyone who runs software, an app or a cloud account, which is most of you. Leo, start us off: what does this look like in practice?
Leo Brandt AI analyst (guest)
Happily! Misconfiguration is rarely a clever bug. It's a door that shipped unlocked. A default account nobody changed, a feature nobody needed but nobody switched off, a permission far wider than the job. Identity is the cloud perimeter, so these settings are your real front gate. And defaults are policy: whoever picked the default owns a slice of the problem, then the humans who never revisited it own the rest.
Leo Brandt AI analyst (guest)
I'd like to say public storage buckets are a solved problem. I'd also like a pony. Default first, humans second, and the bucket is still somehow open to the whole internet.
Maya Okafor AI analyst
The XXE entry is my favourite example, because it's a bug class that lives in configuration. An XML parser is built to follow references, and if external entity processing is left on, a crafted document can make it fetch things the developer never intended. Nobody wrote a bad line of code. They accepted a permissive default. A complete fix is turning that feature off in the parser settings, not filtering the input and hoping.
Maya Okafor AI analyst
XXE goes on my tally of bug classes that should have died in two thousand four. The tally is not shrinking. It's getting a second page.
Jordan Pike AI host
Fine, but what do attackers do with this? Keep it general.
Leo Brandt AI analyst (guest)
As a general pattern, they skip the clever stuff and look for the unlocked door. Default accounts are easy to guess, forgotten features give extra surface, and an over-broad role turns one foothold into many. Static secrets embedded in code, configs or pipelines are the same story: a leaked one works until someone rotates it, maybe never. Short-lived credentials expire on their own, so a stolen one is a much smaller prize.
Maya Okafor AI analyst
Here's where I'd push back on the pure automation story. Hardening baselines are great, but a baseline for your servers won't tell you how your application's XML parser is configured. Check the code and library settings too, and look for sibling issues: every place the app parses untrusted documents. And any quick mitigation is a bridge, so give it an end date.
Leo Brandt AI analyst (guest)
Fair, but I'd start with automation precisely because it's repeatable. Hand-tuned settings drift. Put the secure configuration in your templates, and every new system starts hardened instead of relying on someone remembering. Then your app checks catch what the baseline can't see. Both, in that order.
Jordan Pike AI host
So: Monday morning?
Leo Brandt AI analyst (guest)
Inventory your default accounts and your static secrets, and retire what you can. Where you can, move to identity federation or short-lived credentials, and make hardening a repeatable pipeline step rather than a checklist someone ticks once.
Maya Okafor AI analyst
Find every place your software parses XML from outside, confirm external entity handling is switched off, and add a test that fails if someone turns it back on. Then ask what other unneeded features are enabled in the libraries you ship.
Jordan Pike AI host
Short version: turn off what you don't need, change what came by default, automate it so it stays that way, and stop leaving long-lived secrets lying around. That's the second most common way things go wrong, according to OWASP.





