Docker

yeah i think this might be why a lot of um companies like data dog and snowflake and um what is the other one that's like data centric similar to data dog um new relic maybe or grafana um they've been experiencing a big boom lately um uh when it comes to ai because I suspect it's having all this logging and telemetry and other things that are really useful for understanding how your system's behaving as extra context to feed that humans can't just quickly read.

Same thing for like... very quickly if I try to get my Docker network to come up and something happens.

The error might not be very obvious, but I can say, Claude, can you check and see why it didn't work? And it'll look through all system logs faster than I could type it and go, ah, I can't see.

It's because of this particular issue here.

Now, look at what the defenses were doing at that moment.

Docker sandbox.

The malicious code ran before it.

Environment sanitizer.

Now, look at what the defenses were doing at that moment.

Docker sandbox, the malicious code ran before it.

Environment sanitizer, the process had already inherited the original environment.

Tool allow list, the model never requested a tool.

So what are you doing to secure these sandboxes? I think that's the first question we got to start with.

yeah let's let's do it uh and just know that you know the the acquisition is now about five days old uh but we have been working with black soul we've been a customer of black soul ourselves for rl rollouts and some of the agentic use cases and uh and that's been some of our first questions as well look when it comes to sandboxes people mean different things as it could be as simple as you bring up a Docker container, run some code on it, and then kill it.

And that really doesn't give you the kind of security boundary that you need.

What does are VMs or micro VMs, which is, you don't see every sandbox provider actually use.

Now, look at what the defenses were doing at that moment.

Docker sandbox, the malicious code ran before it.

Environment sanitizer, the process had already inherited the original environment.

Tool allow list, the model never requested a tool.

Now, look at what the defenses were doing at that moment.

Docker sandbox.

The malicious code ran before it.

Environment sanitizer.

That's my plan.

And today we are revisiting Docker.

Oh,

good,

Docker.

It's been a while, yeah? I think it was before I joined.

Yes, last episode was in November of 2022.

That's episode 45, if you want to look it up.

So-

Even on the stable one, it was always the same question, you know, on the all interviews I was getting for the Docker, never put it to the latest one, always ping the version and so on.

Yeah.

Yeah.

So I do think that you should, uh, consider it's not trustable because, yes, there are vulnerabilities in packages, but your own developers can introduce issues, so there will be always something in there.

Obviously, you should put into the effort to re-reduce and minimize the number of possible issues to prevent the intrusions, but you just should think of it that it is vulnerable, someone will get in, so you should not rely on just patching and actually architect around that.

So for instance, you're running in Docker, use, uh, scratch images.

Like for instance, you're running, like, Golang applications.

Fatal flaw.

So OXCC has got an article out about how any user process can escalate to root in the way that Omaki does its default Docker configuration.

So basically, you know, details here how you can actually do it.

But, yeah, the short and sweet of it is that, you know, if you run Docker on Omaki, it can break out and become root on the desktop.

Oh, nice.

Yeah.

I was going to say, surely the pricing will be more expensive, but no, it reckons here that a cloud-run instance with one vCPU and one gig of memory running continuously for 30 days would only cost $570.

Yep.