OpenSSF Operation SLSA Video Series, 16 Nov 2021

Previous Meeting

⏯

youtube image

►

From YouTube: Operation SLSA: Episode 2

Description

These supply chains are getting a little bit of help, thanks to the SLSA Agents!

Music License: MB01SKRLANUVOEN

A

A

Okay, everyone, here's what I found so one of the easiest ways to insert malware into your code is to submit bad code to your repository. You can think of this like trying to submit a forged piece of art into an auction. So obviously the forger wants to pass off their art as the real thing, so in the real world. How would we prevent this from happening?

A

We might get an expert to come and check the piece of art to make sure it's authentic before we allow it into the auction even better than having one expert check it out would be to have two experts check it out. We might also ask for provenance for the piece of artwork itself make sure that the owner who owns the artwork is legit and that we can trace it back all the way to the painter themselves.

A

We can apply the same concept to our code. If we want to stop bad code from entering a repository, we can similarly request one or two experts to look at the code itself before it is allowed into the repository.

A

If we already have bad code in our repository, we need some way of auditing it to find it in the future. If something bad happens, this is where higher salsa levels come into play. Higher salsa levels recommend storing all of your code in a version control system like git or mercurial, so you can keep track of every change that is made in your code base. Each change should be verifiable so that you can see who made the change who approved it and how it was merged into the code base.

A

You should also have at least one to two reviewers for every single code. Change before changes are allowed agent picante. What did you find.

B

I don't care how many experts you have verifying your art. If you leave your warehouse unlocked, anyone can go in and change whatever they want. This is true for software supply chain. You have to lock down your source control repository, so no one can get in and mess around with it. You can't have your source code repository running on someone's computer sitting under their desk. Once your source code repository is compromised, it doesn't matter how many checks you have in place.

B

Someone can go in and change whatever they want without you, knowing using strong passwords using two-factor, auth, strong, two-factor auth, like hardware tokens and securing your source code machine or using a trusted provider can go a long way in securing the supply chain.

A

Fill us in agent verde.

C

We need to talk about dependencies. Dependencies are added all throughout the software supply chain. They can be added and removed without any sort of notice whatsoever. It's like when I was at lunch. The other day there was a dead flower on the table. Presumably nothing changed to the restaurant, but something changed. The wholesaler changed their suppliers down the line changed they didn't catch it. At the end of the day they had a dead flower.

C

What we have is an opportunity for any sort of attack to occur by a malicious package getting substituted throughout the supply chain. Someone could come in take over an open source project. They could go ahead and compromise the package. Someone could make an obscure code change to a package that depends on another package that depends on another package and it won't get detected.

C

The only real mitigation we have is to be very diligent in logging, every one of those packages and what we're using by us recording exactly what's used and where it's used. We can start to look for changes. We can look for anomalies as they occur. If something is known to be a bad package, we can do an audit to determine what is actually compromised in our software supply chain.

C

So it's not surprising that as we keep depending on more and more and more packages, this becomes more and more important in the salsa framework. It's a complicated problem and it's a higher level that we need to achieve. But it's something that we should always strive to be doing so that we mitigate any potential attack.

A

Agent pico: what do you have.

D

So, building on what agent verde was telling us malicious dependencies are what keeps security experts up at night. So when I was in the restaurant, I ordered a salad and my salad came out and I was going to eat it. I looked away for one second and another dish appeared. It looked delicious.

D

It was a hamburger and cake, so this is similar to a software supply chain attack that actually can happen after the build system builds a package, so an artifact can be ingested, bypassing a cicd system and slipped into the supply chain, resulting in an unwanted piece of code that people are using and can lead to a breach.

D

So one way that we can defend against an attack like this is using a framework like salsa, which has a strong tie into the providence or the metadata that's created when you build an artifact. So when we look at this providence data, we can verify it. We can. We can set policies around the build system. We can ensure that we are only accepting artifacts and software packages that are built by particular build systems.

D

And if we see, if there's an unwanted package in there, we'll be able to verify that this package shouldn't be in our supply chain and we will know not to put it in our production systems.

A

Agent roja: what do you have.

E

Look agent pico, that's all fine, but people are putting blind trust in their package managers. What we need is to have transparent audit trails of where those artifacts are being built so that people are delivered what they intend. When I got to the art gallery, I saw fake art pieces held up with real labels that just didn't line up with what people were seeing and what we need is to apply this to software salsa can help us get there.

E

What we need is provenance to be shipped with an artifact so that, when package managers see those artifacts, they can verify the provenance and then make things public to end users. End users should be able to guarantee that they are receiving what they actually intended.

A

So we know there have been many attacks at the restaurant, the museum and the art gallery. Do we have any idea who's doing this.

A