are you affected?
- is your sshd running on musl? (alpine, openwrt, buildroot, yocto)
- is it not using pam? alpine’s openssh-server doesn’t by default, or maybe you set UsePAM no
- are password logins still on? they are by default
- is there an account with just * as its password? key-only users, ldap users, root on alpine
- and some other account with a real password?
- then that * account takes any password between 257 and 1024 characters
on alpine’s official aws and azure images that’s the default alpine user, and it has passwordless doas. so it’s root
on debian, ubuntu, fedora or anything else with glibc? going through pam? dropbear? openbsd? you’re fine
I’m no cybersecurity expert, but from what I’ve heard from those who are, this is the new normal as anyone who formerly had no chance due to being a “script kiddie” can now just prompt an AI* to eventually find a vulnerability.
* oversimplification redacted
Yep, those new fangled slop-kitties. The worst part is that there brute force hacking is being subsidized by people 401ks.
The irony - of course - is that their 401k money goes right into hacking those same people’s 401ks.
… but at least we can brainrot an entire generation with glass fruit slices and put artists out of work?
Now if you’ll excuse me, I have to go convince someone that the TikTok deepfakes they’re watching aren’t real and that cops don’t really employ nanny-boo-boos to unsettle criminals during interrogations. No, I’m not joking.
The way it works is that finding vulnerabilities required a lot of technical understanding and mental effort. A software could have a glaring hole in it for years without anyone finding out simply due to nobody really looking. AI just gives you the ability to do rudimentary exploit analysis around the clock without having to spend your own time. This is a wake up call to anyone making software in sloppy fashion.
But it isn’t quite as easy as saying “AI hack into this service”, if the person doing the hacking is using the same model as the person who made the service, and that person already asked the AI to do that and fixed the resulting exploit prior to release.
Even projects that use AI in a limited fashion to code or even not at all, often use AI to at least do some red team stuff prior to release.
I don’t really foresee any way this doesn’t even out eventually.
It’s not quite as simple as that. The author in this case followed a pretty specific trail to find the bug.
Where AI helped was as the last step. They already knew what the bug would do and where to look, but actually looking through the source code for all the possible musl implementations out there would have taken a lot of time.
So they brute-forced it instead and asked AI to spin up actual running instances of all those implementations and simply tried the bug against each of them, rather than looking at all the code.
Fair enough, but the overall idea stands:
The bar has been significantly lowered for what it takes to hack.
That’s because it turns out there are tons of vulnerabilities out there introduced by very lazy and/or dumb code.
It isn’t that AI has lowered the bar for hacking, it’s that it’s showing us just how much bad code is out there. The bar was already on the ground, we just didn’t know it.
After all is said and done AI will actually end up raising the bar for security everywhere.
Unfortunately we have a lot of legacy crap to clean first.
You’re doing mental gymnastics to argue with me. Let’s try what I said worded differently in a way that hopefully helps you realize we agree:
It is now easier to hack.