How Project Akrites Is Building a Global Open Source Security Response Team | CRob, OpenSSF

AI powered tools have made it extremely easy to find vulnerabilities in open source code. What used to take security researchers weeks of manual work now takes an AI tool minutes. And every organization is doing that. They're scanning the same open source packages independently. And more people who know about these unpatched vulnerabilities before the fake ships, the higher the odds that it leaks to attackers before it is patched. The irony is a bank and a hospital may be using the same obscure open source library. They may find those vulnerability, but they don't talk to each other and they won't even know unless or until it's exploited. And to address this challenge, the Linux Foundation is announcing a new project called Equities. It's kind of a shared coordination layer sitting between the people who find vulnerabilities and the open source maintainers who actually have to fix them. That's my understanding of what this project is all about. And today we have with us, once again, Christopher Robinson or Crob, CTO of Open SSF and chief security architect at the Linux Foundation, to walk us through it. First of all, Crob, it's great to have you back on the show. And tell me how close or far I was from the real problem that we are facing due to open source and AI fighting against each other. Sometimes I feel that, "Hey, that's a good thing because AI is finding a lot of vulnerabilities which we could not find otherwise." But at the same time, the volume, the lack of coordination is creating a lot of challenges. So, first of all, correct me how far I was from the reality. >> Oh, thank you for having me. And you were actually pretty close. It was around the fall of last year where the community really recognized that there was this increased use of these LLM's and other AI enabled tools and we had a lot of conversations in the community and across the ecosystem about AI slop. And it was the case that last year uh many people were using these tools and they were just throwing reports over the walls at uh upstream maintainers and corporations and the quality to be perfectly blunt was pretty terrible. I and there was this huge effort and the Open SSF was talking about in part of our vulnerability disclosure working group, let's write some guidance for researchers and for projects to try to help mitigate this AI slop problem. Around the beginning of this year, January, February, something changed in those advanced models and the reports actually started to get pretty good. And this this isn't universally they they aren't a universally beloved thing, but generally the quality and uh the data you got out of these types of tools was much much better and like thinking about like where we are today in the summer uh the the reports are almost as good as if you had a professional kind of seasoned security researcher kind of applying their time and you're absolutely right. It used to take researchers weeks, months, uh sometimes maybe a year to kind of think through and figure out how to present this vulnerability and to get it fixed. And the time now we have with these um automated tools like a using LLM's, whether it's an advanced kind of frontier style model like a mythos style thing or even if it's an open weight open source model like a Kimmy where we're seeing very good results with these open weight models and the time has been compressed where now we have hours after a public model finds something before potentially somebody else could discover it. You were very close. >> So, thank you. Now, talk a bit about first of all, how do you say Equities and what is this project all about? What is the idea behind the project and how it's going to address this problem? >> So, the project name is Project Equities and you can go to equities.org to kind of see our public information. And as we move forward and start executing on the program, you'll see more information both on the website and then our like our GitHub repositories and put more public communications. Um and you what we've noticed is there are a lot of people, whether you are you know, somebody that has a $20 chat GPT or a cloud code account and you want to do scanning or you're a very large organization that might have access to some of these larger things where there's been a a substantial amount of stakeholders that have been trying to find and figure out what to do with all of these potential vulnerability reports. And that's where a project Equities stands in. Um the Equitai were a group of Byzantine soldiers that used to live out in the outskirts of the Byzantium Empire and they were kind of like an early warning system. They would keep an eye on things. They were defenders of the frontier and then as invaders were coming or if there was trouble, they would be the ones that would help uh try to address that particular problem, but then make sure that everybody back in the empire was warned. And that's where that kind of thinking that we're at is again, there were a lot of people looking to try to find these problems, but very few were thinking about the problem of what do I do with all these findings? How do I get them upstream? And that again, thinking about last year with the the reaction of upstream to AI slop is the worst possible case is somebody runs a scanner and we start filing thousands of reports without arming upstream so that they can understand how to do what the report is, how they can take action, and kind of giving them um additional tools and resources so that they can address that those findings in something that's uh acceptable within that upstream community. >> And also talk a little bit about uh why launch this now. Was there any specific moment or catalyst that pushed this from kind of an idea, a thought process to a reality? >> Mhm. Well, if you've been around the OpenSSF for any period of time, people the long-time players at home might remember we had a thing called the mobilization plan about 4 or 5 years ago. And actually one of the work streams that was identified by our community and the board was this incident response team. And at the time, the problem wasn't as pressing or wasn't as obvious to folks, and that didn't move forward. But this year, um starting off with efforts like Anthropic's uh glass wing program, what um OpenAI has done with uh GPT cyber and these other programs is now people have access to these very capable advanced models. And these tools can do much more sophisticated work than your traditional security scanner can do. And we started to start um those organizations started to recognize that they were finding dozens, hundreds, potentially thousands of findings on projects across the open source ecosystem. And again, we've all joined together to try to make sure that those are uh responsibly disclosed, and making sure that they're empathetic to the upstream project, and then most importantly, that there is some type of public treatment so that when these issues get uh released as general availability, public disclosure, that everybody that consumes that package has access to those fixes as quickly as possible. >> If you look at some of these major open source projects that literally are people may not realize backbone and foundation of modern economy. Linux kernel, you know that or Kubernetes, there's so many open source projects. What does AI and the security of these projects mean in this new world where vulnerabilities can be found easily and now the problem is that both good guys and bad guys are looking at it and I always say as good guys you have to be right 101% time. Bad guys have to be right only once. So how do you look at this word? >> And from our perspective, obviously open source is deeply ingrained in our DNA and we feel that this is the most effective model of development and co-engineering in existence. And if you look at these frontier model systems with the exception of like some of the hardware components and then some of the like the training information, all of these things are deeply deeply dependent upon open source libraries and communities for their commercial offerings. So it really depends on who you are and kind of what resources you have access to. There are amazing for fee paid services that these models have some great capabilities, but thinking about open weight and open model models, we feel that that is a a great approach that people can effectively leverage because you can understand much better. Just like open source code where you can understand how this model was developed, how it was trained, how the decisions are made with the weights. So this is a a path that we think is just as important and while the open weight models and these the newer open source things might not have all of some of the advanced capabilities like some of these frontier models have are very sophisticated in their ability to do vulnerability chaining, you can do that with, you know, less cost models. You don't have to go with the most expensive thing on the planet, but you also get very good results. And what we've seen in our testing and engagement is that the open weight stuff actually we we kind of prefer because we have better ability to steer how that how that information is trained and then share that and then people can look at the decisions we made or as we were putting the model together and that's where you can have like this expert system that's trained to think just like me and I can have that conversation and not worry about disclosing confidential information to a potentially public resource with other people share. >> Excellent. Thank you. Now, I have been tracking of course OpenSSF since early days and a lot of security-related projects that predates OpenSSF same way you have been tracking them. Talk a bit about how is this project Equities different from OpenSSF? Is it complementing it? Is this Is it competing? I mean, there is no competition. It's always cooperation in open source world, but I just want to understand the the the existence of these these different projects. Are the goals different or the goals are same? Is new problems emerging? So, you have to come up with more focused projects. So, I just just talk about that. >> It's very definitely complimentary efforts. The OpenSSF has it'll we're coming up on our sixth anniversary. We have a six-year track record of delivering best practices, guidance, documentation, training, public policy engagement and those are things that we'll continue. And you can think of OpenSSL OpenSSF along the lines of kind of a SDLC, a best practices and secure by design focus. Acquities is laser-focused on one piece of the SDLC and that is the let end phases where you're in maintenance and updates. Laser-focused on finding the vulnerabilities, making sure that those vulnerabilities are accurately documented, and then that those reports are passed up to some type of upstream project for determination on how it goes out the door. So again, it's very complimentary. We have some work within the Open SSF that talks about this piece, our vulnerability disclosure working group. And ultimately, what the plan is is as the Equities CERT, the security incident response team, starts getting in there and working with upstream projects and across the ecosystem, is we're going to be cataloging these good practices and then passing them back over to the Open SSF for creation of a white paper or training, kind of whatever the particular, you know, artifact is needed. And again, kind of leaning into that subject matter expertise of that that community as opposed to incident response, which is kind of a a very small slice of the world. >> And uh as I was saying in the very beginning that it is kind of neutral common place where, you know, no matter who can uh so if organizations do want to get involved, should they get involved with Open SSF or should they get involved with Equities? Uh because sometimes it may lead to confusion. Hey, who should we talk to? What about you? So give us some guidance on who they should talk to and what does the membership involvement looks like there? >> It depends on which kind of story resonates with the particular stakeholder. So if you are interested in that long-term sustainability, if you're interested in tooling and training and education to help the supply chain, Open SSF is the place to go. If you feel that you have security engineering, like penetration testing, red team, product security experience, and you want to help with this particular problem and helping set some of these norms and helping uh broadly upstream coordination has been a challenge for the industry. And this is the first effort. But it's So, if you're interested in focusing on that specific piece of the problem, that's where you can onboard is going through the Equities project. The Equities, we have a couple different membership levels and broadly the core tenant is that if you're going to participate within this group, you not only need to have, you know, pay your membership dues, but there's also a requirement of dedicating subject matter experts, engineers to the effort. And that's where the idea is we have we'll have a core team of incident responders and we'll build this coalition of red and experts from across the ecosystem so that when we need assistance or maybe we have your engineers have specific relationships with critical projects, for example. It'd be great to have you in there participating trying to help us help that project specifically. Or if you had more generalized experiences. But again, if you're looking to do, you know, the analysis of vulnerability reports and then kind of the negotiation, communication, and disclosure of those, that's one particular set of skills. But if you're more interested in the kind of that long-term sustainability, helping set standards like Salsa, which is the kind of the de facto supply chain security standard, that's more appropriate for the Open SSF where the Equities will not be doing that type of longer-term sustainability type things. >> You and I were crazy about CRA back then, you know, and we always talk about that. Because of AI, that is also becoming a big problem from if I ask you to wear your CRA advocacy hat for a few seconds, how is this project going to help with that? >> Well, I The theory is that if we're able to get these vulnerabilities discovered and then have fixes distributed upstream more quickly, hopefully that will help all of the manufacturers, which are the ones supplying the products to the European citizens on the European market. So, the theory is that by doing this, we're able to help them meet their obligations. And then, you know, as we're engaged with upstream projects, some of these projects um might not have a steward or you know, a big brother or big sister to help. Part of our mission is to as we discover these projects that either, you know, have retired, they might not be operating upstream, they could be end of support, or they might need help. Will be part of the CERT's mission will be to try to find stewards to help support these projects long-term. And that's where again, it's um it does not directly help an organization kind of achieve their compliance needs for the law, but in theory, we're able to get data out there more quickly so that they those organizations can be more responsive. And then, you know, as they have capabilities, we're trying to encourage those members to step up and become stewards and provide that more long-term sustainability for these projects. >> What kind of involvement, engagement, uh you are seeing you also have LF Europe now as well. You also have Linux Foundation India as well. Talk a bit about what kind of engagement are you looking at from global perspective? >> So, from our perspective, open source is truly a global international effort. You have people contribute from every location on the planet. So, to serve just one jurisdiction does not help because very rarely are projects kind of defined within a particular geography. And so, our goal is to try to be this neutral international body to help the open source. And very broadly, we've had great conversations with international governments. We're in conversations with the US government and like their Gold Eagle project and we've been long-time partners and contributors with CISA, NIST, and MITRE. So, that is again, part of what we're doing. But, through our work with the CRA, I have excellent engagement with our friends at ENISA and the European Commission. And you know, those are stakeholders that we've been briefing about the project and you know, we feel that as these issues go public, we want to make sure that that information is shared at a global scale and not restricted to a particular geography. And we're also in conversations with our friends like for METI in Japan, talking with folks in India, Korea, Australia, UK. So, we we were thinking the program is designed to be as you know, international and globally cooperative as possible. And you know, my team will be an international team. I've already got folks from several different nationalities participating and that's kind of our goal is that at eventually when the team is in full operation, I have kind of a global follow the sun approach where I'll have experts in many different geographies that'll be available to assist depending on what the particular issue arises. >> Rob, once again, thank you so much for joining us and walking us through what this project is all about. And of course, those who are watching, please head over to acrtis.org and check out the latest work and please get involved. That's the only way of how open source works. Thank you again for your time again and I will see you in the next one.