41 minSeason 2, episode 32

The STRIDE Controversy: Evolution vs. Extinction in Security Models

With Chris Romeo, Izar Tarandach, and Matt Coles

Threat modeling & security architecture

We discuss a controversial LinkedIn post claiming "Threat Modeling is Dead." While the STRIDE methodology may need updating, it remains a valuable "gateway" tool for teaching security concepts to developers without security backgrounds.

Episode title image: The STRIDE Controversy: Evolution vs. Extinction in Security Models

Listen to the conversation

About this episode

We discuss a controversial LinkedIn post claiming "Threat Modeling is Dead." While the STRIDE methodology may need updating, it remains a valuable "gateway" tool for teaching security concepts to developers without security backgrounds. We discuss how STRIDE serves as a useful categorization system, emphasize that dogmatic approaches to threat modeling are problematic, and argue that what matters most are results rather than strict adherence to any particular methodology. Our conclusion; STRIDE is still alive and relevant, but it could benefit from an update to demonstrate its continued applicability.

Follow AI Security Table:

➜ Home: https://securitytable.ai/

➜ X: https://x.com/AISecurityTable

➜ LinkedIn: https://www.linkedin.com/company/ai-security-table/

➜ YouTube: https://www.youtube.com/@AISecurityTable

Transcript · 41 min conversation

0:09Izar Tarandach Sponsor us, you will see me drinking with my finger up.

0:15Chris Romeo So how many espressos do you drink a day now?

0:19Izar Tarandach How many are possible?

0:23Chris Romeo So this episode of Security Table brought to you by Isar's espresso machine, which I don't know what brand it is. But it appears to be making him very happy, and it can make you very happy too. And then Matt was showing a slightly bigger cup.

0:37Matt Coles Wait, slightly bigger?

0:38Chris Romeo And look at these. Ours is like a mini. No, his is like half the size of that.

0:42Izar Tarandach It's not the cup, it's the coffee.

0:45Matt Coles It's the, it's the, it's the, it's tapered. Go on, it's smaller. Look, he's, he's drinking it properly too with his finger.

0:55Chris Romeo It's so funny to watch him. He's like so distinguished. He's like This is, this is good, old chap.

1:01Izar Tarandach For distinguished, I have to go with my monocle.

1:05Chris Romeo I mean, yeah, the pinky has to go all the way out at that point. $1 million.

1:15Matt Coles Just don't do that while you're drinking.

1:17Chris Romeo Be funny if you could do both. See if you can combine both together. Okay, Quell, enough fun with this. We do have a purpose. We are here for a reason. And we're going to talk about a post somebody made on LinkedIn called Threat Modeling is Dead. Let me just read it. Why don't I read it to set the stage? It's shorter.

1:40Izar Tarandach Was it threat modeling that was dead?

1:41Chris Romeo That's what it says.

1:42Izar Tarandach I'm going to read it.

1:42Chris Romeo It actually says threat modeling is dead. Direct quote to put this on the record so that the audience knows what we're talking about here. The post in bold says threat modeling is dead. And then Stride has not aged well. I've seen firsthand the thick PDFs and Excel files that my clients have had delivered to them by consultants. Those docs have about a million data flow diagrams, giant tables with columns spanning over multiple pages, and a bunch of unactionable data. Software engineers and hackers both move much faster than in the past. I don't even know what this, I'm not gonna comment. A list of high-level threats is not really useful when it comes to agile secure software development because exploitable attack vectors are usually down in the weeds. It's the detail that's missing from STRIDE reports despite the huge amount of documentation. What can we do instead? ATT&CK modeling. I've been favoring ATT&CK modeling over threat modeling for some time now. It has only 3 steps. Analyze attack surface for each component of the design. Produce an attack register, a list of potential attacks against the exposed attack surface. Produce a list of associated security controls to mitigate the attacks. ATT&CK modeling cuts the documentation significantly and produces relevant actionable data with none of the fluff. Another bonus, the attack register can be used as a scope for the pen test. And that is the setting. Where to begin?

3:05Izar Tarandach Where to begin?

3:10Matt Coles I don't know where you do. Where do you want to begin with that? Because there's so many, so much wrong with that statement.

3:14Izar Tarandach Wait, wait, wait, wait. I think that I'm personally going to start with, with Acadia. Because I did put a comment there in that post. And I have in the past in this here podium and podcast declared that DAST is dead. Exactly. At the same time when we did declare that, we did this coming from places where we are users of it and we have seen it in many places and we have seen the results and how they're used. And at the same time, in that same episode and many times after, we did say that it still has its uses. So we went for it's dead with a bit of a flair for the drama. But in this year post, my comment there was that I would love for people who declare things to be dead to at least take their polls beforehand, because I think that stepping out and saying, hey, one of the things that's like the most popular and that people use as a huge onboarding tool to get behind this chat modeling thing is that when so many people are actually using it every day and making good use of it and teaching it and learning it and I don't know. I think that perhaps we should be a bit more careful with the things that we declared dead. Could just be my age talking, but perhaps check before.

4:53Chris Romeo Well, he's saying threat modeling is dead. That's a pretty big statement.

4:59Matt Coles It is. And then on top of that, to go ahead and actually suggest an alternative threat modeling approach as a replacement.

5:07Izar Tarandach Look, it's dead clearly because we already solved all the possible problems. So there's no more need for truck modeling. Everybody's doing the right thing. So yeah, clearly it's dead.

5:17Chris Romeo Yeah.

5:18Matt Coles AI's gonna solve everything, right? I mean, come on.

5:20Izar Tarandach AI does—

5:20Matt Coles No, AI won't solve everything.

5:21Izar Tarandach AI already did solve.

5:23Chris Romeo So I mean, let's, let's, let's pick this apart a little bit more. So he says Stride has not aged well. Okay, great. That's your opinion. We can, we can leave that on the table. But then he says, I've seen firsthand the, the PDFs and Excels delivered by consultants with million, a million, about a million data flow diagrams. That's probably an exaggeration. I don't think anyone ever sent a million data flow diagrams in a document. I kept giant tables with columns and a bunch of unactionable data. I, I, I got a, an idea for you here. You might be using the wrong consultant if that's the data that they're sending to you. Like, you could apply that to anything. If someone's doing a home inspection for a new home I'm buying and they drive up with a pickup truck with 1,400 boxes in it and unload them and say, that's your inspection report, I think I'm going to probably find a new home inspector, right? Like, I'm not going to parse through 1,400 boxes of paper to see what you think was wrong with the house I was going to buy.

6:19Izar Tarandach You forgot the most important part. The guy driving with the 14 boxes has a cousin who's a contractor and can fix each and every one of them. Okay? So I think that Albert Anhai has a great comment in the post where where he points out that there is such a thing as, uh, what's it that he call it, the consultant, uh, industrial complex.

6:42Matt Coles Mhm.

6:44Izar Tarandach And this person's clearly a victim of that.

6:47Matt Coles I mean, you've talked about, you've talked about it for, for years now about, you know, having a consultant parachute in, give, give a bunch of stuff and leave, you know, each just the seagull, the seagull consult.

7:01Izar Tarandach Consultant, they swoop in, shit on the table, leaves, and you have to clean that mess.

7:09Matt Coles And so if you're paying your consultants by, you know, but for the words, then, you know, you're gonna get what you get. Uh, but, but I mean, that's a, that's a, that's a using— was it hyperbole for, for opening of argument? Uh, is kind of, is very disingenuous. So starting with threat modeling is dead, STRIDE is not aged well. To, to hide, to say that, so first off, there's probably some confusion about what STRIDE is versus what threat modeling is, right? STRIDE, of course, is a, is a, has been seen, I think, or has been popularized as a methodology for performing threat modeling, which relies on DFDs. And in practice, that's the way it's been implemented. STRIDE on its own is really just a categorization of threats and that there's a process of looking at whatever diagrams or models or descriptions of a system you have to match those things to threats, to the threats in STRIDE.

8:18Chris Romeo Yeah.

8:18Matt Coles And, and so to say that threat modeling is dead and STRIDE does not age well, if you're using STRIDE to categorize threats, then you're still doing threat modeling. And, and I guess the other, I guess the other important part here is to consider that what we call threat modeling, for people who are not watching this or who are listening to this, go to YouTube and watch this, this video. So what we call threat modeling, I think, has, has evolved over time. Traditionally, threat modeling was looking at threats, purely threats. What was somebody going to, what is somebody going to do to your system? More importantly, what is somebody looking to do to your system? Meaning what impact or effect do they want to have, right? Do they want to do site defacement? Do they want to, you know, spoof your data? Do they want to sniff your data on the wire? What threats exist? But over time, I think that has evolved and the term threat modeling is now used to cover, as an umbrella term to cover things like looking at threats and attacks. And of course, so what is an attacker going to do to cause a threat to occur or to, to implement a threat? And what weaknesses or vulnerabilities are they going to leverage to make those attacks possible? And so threat modeling looks at weaknesses today. Threat modeling looks at weaknesses, vulnerabilities, attacks, and threats. Informs risk. We've talked about this at length in the past, right, about informing risk. And so threat modeling as an activity, I'll even go on a limb, and I know some of our audience will be happy about this, is we should not use threat modeling as a term. Maybe just, we should call it secure design review and be done with it.

10:05Izar Tarandach Will we?

10:06Chris Romeo Please no.

10:07Izar Tarandach Review?

10:08Chris Romeo Yeah, review is a different— secure design is one thing. Secure design review sounds like a crappy meeting I got to sit in for 2 hours where too many people want to talk.

10:17Matt Coles Sure.

10:18Izar Tarandach Come up with a better term.

10:19Matt Coles Come up with a better term if you want, but that's fine. But threat modeling is fine.

10:22Izar Tarandach Security design roast.

10:24Chris Romeo Yes. The roast of your secure design.

10:28Matt Coles Critical design analysis.

10:31Chris Romeo Is that what you said? We come up with microphones and we say mean things about you and your design. I love this idea.

10:35Matt Coles Shall we call it a critical design analysis? And that maybe is a better way. Critical? That sounds so crazy. It does sound too critical.

10:42Izar Tarandach Things we would have liked you to think before you did them.

10:46Matt Coles The T, yeah, put that in after number 4.

10:48Chris Romeo The T, B, C, C before. Can't even think of what it is, but coming back around to Stride though. So I may have an, I may, I don't know, I may have a slightly unpopular opinion of how I use Stride that you guys, I don't know if you guys are going to agree with this or not, but to me Stride is a gateway. Not gateway computer, because that's from a long time ago.

11:10Izar Tarandach Yeah, we talked about that.

11:10Matt Coles Not a gateway drug. With Kestrel.

11:12Chris Romeo It's not a gateway drug either, but it's a gateway to threat modeling, right? And the way I've always approached it is Stride is a way to teach people to, in the very early days, to threat model when they don't have the security knowledge experience. I think what this post is missing is the reason why Stride exists. You can't take this attack modeling idea and hand it to somebody, a developer who doesn't have security background. They don't have all that stored in their head. And using their, your scanning tool as the source for the things they need to think about, why the heck do I need to do that? Why don't I just run the scanning tool? Why don't I just run Burp with its different attacks?

11:48Matt Coles Because you have to have code already that's already functioning and you're really late. Wait, well, let's not go back to that episode about you're really late in the process.

11:57Izar Tarandach But wait, wait, wait, there is one thing here that's important. I think that Okay, no, I don't think, I agree with you that STRIDE is a gateway as are other things, right? But that brings me to a point that I keep not understanding. Why do threat modeling tools insist in labeling things with STRIDE?

12:19Matt Coles Because it's a categorization system.

12:21Izar Tarandach But again, you already have the threat. Why is it so important now?

12:25Matt Coles Because it's recognizable and it's understandable to those who are gonna see reports.

12:29Izar Tarandach But is it actionable or is it just one more field in the report that I'm looking at and think, oh nice, this is spoof. Yeah, I know it's a spoof, but the threat is saying this thing spoofs this thing.

12:40Chris Romeo I agree with you. The STRIDE is a list of categories of threats that stand on their own. They don't define anything else. Like the way we approach the Devicis, like STRIDE is a set of threats that exist in the codex. But we also have threats from lots of other categories. I don't go through and tag things with stride pointers because I will disagree with you.

13:06Matt Coles I will, we can agree to disagree.

13:08Chris Romeo No, no, no, you gotta make, hold on, you haven't said why you disagree.

13:10Matt Coles Go ahead. Well, because there's a whole range of threats that constitute spoofing or tampering, right? They are themselves categories. If you look at the early implementation of the Microsoft Threat Hunting tool, Right? They didn't have one threat that said spoofing. They had a bunch of threats that were under the spoofing category.

13:28Chris Romeo But that constrains you too much today in the modern world where we have literally thousands of possible threats.

13:35Matt Coles And by the way—

13:36Izar Tarandach but imagine—

13:36Matt Coles and by the way, threats don't have to be singularly categorized, meaning a single threat may not be one STRIDE category. There may be multiple STRIDE categories baked into a single threat.

13:47Chris Romeo Yeah, but then you're using STRIDE too much as like the end-all and be-all.

13:51Izar Tarandach No. Wait, wait, wait, wait, wait, wait, wait, wait, wait.

13:52Matt Coles How did I say that? Well, I think you're putting words in my mouth here.

13:57Chris Romeo Well, of course I am. That's how I work.

14:00Izar Tarandach Let me invite you to look at this in a bit of a different light. If you look at the CWEs and you start from the root and you go down, things at the end of the day, when you identify threats, they get much more useful when you reach the final leaves than when you play in the middle.

14:19Chris Romeo Sure.

14:20Izar Tarandach STRIDE to me is a middle, is equivalent to a middle layer in the CWE. It's that generic thing.

14:27Chris Romeo It's a class, yeah.

14:27Izar Tarandach All the things that come under this are in this category, right?

14:31Matt Coles Agreed.

14:32Izar Tarandach So if I had tools pointing me to specific CWEs, okay, let me walk back from that back a second. I recognize that there is an educational value to pointing at a certain threat and saying this falls under this category of STRIDE. Because in the future, when the person is doing their own analysis, they are able to remember that and say, oh, wait, this new thing that I found is similar to that old thing that was found, and that was under spoofing. So this is another instance of spoofing, right? So there's an educational element there. But in terms of practical sense of making sense of the weakness that I have in hand, going as low in the CWE scale, not, not, not using the tree as a categorization, but scale of, of, of detail. I think it's still more useful than stopping in the middle at the stride level of categorization.

15:29Matt Coles Agreed. Right. We're not talking about stopping at the middle. We're talk using it as, as a reference point.

15:35Chris Romeo So do we have, so Matt, let me ask you this then. Do I have to have a stride category on every threat?

15:40Matt Coles No. And in fact, I, I think you won't.

15:43Chris Romeo Yes.

15:43Matt Coles Right? So first off, if you have a privacy threat, you're not going to have a strike threat category, right? So not everything is going to map.

15:49Chris Romeo Do you mandate a Linden categorization for privacy threats?

15:56Matt Coles You're putting a word there that is— Kind of interesting.

15:59Izar Tarandach Mandate. We don't do mandate.

16:01Matt Coles We don't do mandate. So when—

16:03Chris Romeo I mean, should you? Okay, so let me rephrase. Your Honor, I hear. Let me rephrase the question. Would you use the Linden categories similar to how you're describing using Stride as a categorization system? So if you had a privacy threat, you would pull through and say you should use the Linden categorization?

16:24Matt Coles Probably highly recommended, yeah. I mean, as far as other privacy threat modeling goes, Linden is the reference point, right? Similarly, you know, if you had, so similarly, if you had, um, I guess, so here's where it gets interesting. You could use other things in place of stride.

16:44Izar Tarandach Yes.

16:45Matt Coles That are also similar reference points that are that similar middle layer of the, that we were talking about. What else?

16:50Chris Romeo Give me another example.

16:51Matt Coles So include, sorry.

16:54Izar Tarandach Cool.

16:55Chris Romeo What an intro. Wait, wait, wait. It's your time now.

16:58Matt Coles Includes no dirt. Or was it SPARRA? But there's other acronym. It's— everyone's following the acronym base, but it's basically a set of principles.

17:09Chris Romeo Nothing generic though. There's nothing. Those are all things that were very threat modeling specific, like NoDIRT was, uh—

17:14Izar Tarandach Principles.

17:15Matt Coles Well, you could go, yeah, you could go back to design principles from like NIST 800-160 Volume 1, or you could go to CWE or KPAC. I mean, not KPAC, because KPAC is really attack patterns and And you have the MITRE ATT&CK framework, which are ATT&CK patterns, but CWE, CWE class, uh, class, weakness classes is another good, another good reference point as if there was already—

17:41Izar Tarandach but wait, I, I want to bring out something very, very important that Matt just said, apart from the other very important stuff that he keeps saying. There are other things. I think that. Something that we see recurring at this phase of the threat modeling community is dogma. There's a lot of dogma out there. You do stuff this way. This is the way that you do stuff. Now, I don't know if it comes from me being lazy. I don't know if it comes from me being nice. I don't know if it comes from me having seen a lot of stuff out there, but I don't care what you use. I care about your results. If you are in the point of your journey where stride gives you the best results, then use stride. If you are in a stage of your journey where I'm sorry for you, but cooking spaghetti does the best for you, go and cook spaghetti. You know?

18:39Matt Coles Well, and so what if I don't like spaghetti?

18:41Chris Romeo Again, I don't like spaghetti. Okay.

18:43Matt Coles Well, then your journal, look like diet.

18:46Izar Tarandach No, we don't like the dogmatic approach to there is only one way of eating and it's only spaghetti and it can only be cooked this certain way and everything else is that. That's what we don't like.

18:59Matt Coles You use a giant fork and spoon and you're— Yeah.

19:02Chris Romeo I want to, I just want to reiterate, I want to, what you said there was really important that I heard Cisco's previous CEO, John Chambers, used to say this all the time. Never confuse hard work with results.

19:14Izar Tarandach Where do you think you heard that first?

19:16Chris Romeo I probably heard it from— I didn't know that. I didn't know you knew him.

19:19Izar Tarandach Nope, nope, nope. Wow.

19:20Chris Romeo But that's what you're saying though, is like, it's— and we could just, we could change some words and say, never allow your threat modeling approach to, you know, never confuse that with results. So if you're getting results, that's the end of the day, that's making the product more secure and more private.

19:39Matt Coles I think there's a threat modeling manifesto principle or value about that, if I recall.

19:45Chris Romeo Sekhna?

19:47Matt Coles So, um, so one, I, oh my gosh. Um, yes. Gosh, gosh, gosh, gosh, begosh. Um, I was going to say something kind of important too. Oh, uh, I wanted to counter, I wanted to counter your, just your thing about you should start with Stride because it's easy. Um, the reason why we have fret libraries is because Stride is not easy. If you take it at an extreme, right? So understanding what a spoofing threat is versus what's an implementation of a spoofing, or what's a, what's an actual threat that is under the category of spoofing. A lot of people start with STRIDE, and that's, I think, where a lot of people get lost unless they have somebody with security knowledge and experience on their team who's able to do that translation, which is why we have threat libraries.

20:36Chris Romeo You're describing threat modeling in a vacuum. Which I don't recommend in the early days for people.

20:41Matt Coles I'm not suggesting that at all, but that's why we have threat libraries, is so that somebody can take it and do it.

20:44Chris Romeo I agree, I like threat libraries, I think they're powerful. I also like people to understand how to threat model by hand, if possible. It's like doing mathematics, right? Like, I'm not a math person, I am terrible at math, but there's always something about when you understand how to do the problem versus how to get the answer, it just changes your perspective on the thing and allows you to build other things on top of that thing because you got that concept, right? And so that's the only caveat for threat libraries. Yes, they're powerful, but I don't want to be in a mindless world where people are just like, I ran the tool, I got these 3 results, I created the tickets and I moved on. I want them to understand because, listen, I build a threat modeling platform and brainstorming has still got to be a part of what we do here. I can't create a threat library that takes Isar and Matt's experience in the different jobs you've had and meld them together into something that will spit out creative new threats as people have ever thought of. I need the human beings to drive that based on their experience. And so that's my only caveat there is I like, I want to make sure people know how to do this versus just knowing how to access a threat library.

21:55Izar Tarandach So that's the thing, threat libraries and What was the other thing? The other thing that we were supposed to— Forget about it. Threat libraries. So my problem with threat libraries is that they lead to two things. First of all, threat modeling by looking at the rearview mirror. Like that's been my analogy for a long time ever since I first encountered threat libraries. You're driving a car and you look at the rearview mirror at the things that already passed. Eventually you are going to hit something new going forward, but you are so focused on that thing that you won't see it coming. You won't recognize that it hits and there's probably going to be a horrible accident. So while I understand their value and I see their value, I think that they are by definition limiting and in double, triple, quadruple airports dangerous for people who are just going on that on-ramp to assume as their methodology or their way to learn. And the other thing that I think they lead to, and, and this is something that I, I have talked with, uh, Brooke extensively about, is, uh, cargo cult mentality. At some point, the threat library becomes a threat checklist. And then moving from there, you start just checking those things because it's the things that we always check because it's the things that when one person checks way past in the past, they got results. So if we keep doing those things the way that that person did, we might get results.

23:34Matt Coles Mm-hmm.

23:35Chris Romeo And, and you call it, hold on, just what you, what, what were those, the two words you used to call this?

23:40Izar Tarandach Cargo cult.

23:41Chris Romeo Cargo cult.

23:42Izar Tarandach That comes from a thing that happened at the end of World War II in the Pacific. They had the US-built airstrips in small islands that had some native population. And the native population after the war, they developed the habit of building runways, like sea miles of runways, because they associated that with sorium. It's going to sound very condescending, which the big bird coming from the sky and bringing food and things. So they believe that if they only build the runway, the big bird will come back. And that's how tough things were.

24:26Matt Coles Yeah. And we should just highlight that, of course, today is Veterans Day here in the US. Oh yeah. Uh, and thank you all. So, um, you know, so what you highlighted though is, is really important and why Stride actually still has value. So if you use a threat library to identify threats and you identify the threats that are already encoded in the threat library, as soon as you incur, as soon as you identify the opportunity for something new, right? Assuming you're not looking at, you know, blinders on looking at only what's in the threat library and, and you have the ability to find or where you do come across something like, hey, this doesn't fit anything in our list, but it's still a threat we want to consider. Well, how are we going to consider that? Well, you may take it up a notch to that parent class, right? AKA the STRIDE classification. Hey, this is like a spoofing threat. It's not in our threat library. It's not a listed threat as a spoofing threat in our threat library, but we've now identified something new. Well, how do we describe it? Well, we might describe it by mapping it to a stride category. And now I have a way, a basis, you have a, now have a basis for describing it in a way and looking for its properties to be able to then put it into your threat library for later, right? So that you can now start building on your threat library.

25:47Chris Romeo You guys ready for some controversy here? Absolutely.

25:49Matt Coles What is, what would today be without controversy?

25:52Chris Romeo This is, I'm going to, wait, I need another coffee. Yes. Take it. We're going to, we'll be back after this, these short messages for Yezar to get an espresso or two. Is the cat— are the categories that are represented in Stride still relevant today like they were? I just, this has got me thinking as we've been talking, I went on this kind of thought experiment. Spoofing and tampering both. The counter to spoofing and tampering is encryption, maybe some input validation. But normally we think of encryption as being the counter to those things. Like, if you encrypt or if you use multifactor authentication from a spoofing perspective, how prevalent is spoofing anymore though?

26:38Matt Coles Well, you're asking the wrong question. It is not complete, but it is still relevant. Spoofing still exists. Just be— first off, if you implement multifactor authentication, you do not eliminate spoofer.

26:50Chris Romeo Of course, but it's a great step in the right direction though. Yeah, absolutely it is.

26:55Matt Coles Uh, Matlocky. And if you implement encryption, whether on the wire or at rest, you're not going to solve all tampering problems, right? Or all information discovery problems.

27:04Chris Romeo Give me an example of a tamper. How are you going to tamper? What are you going to tamper with that's something that's encrypted?

27:12Matt Coles Well, what if I tamper with the encryption key?

27:15Izar Tarandach I can tamper and it won't unencrypt, but it will still break the way that the thing's supposed to work because they did not consider the fact that something can fail to be decrypted.

27:25Chris Romeo Yeah, well, my point was—

27:28Izar Tarandach Before that—

27:28Matt Coles Well, then you may actually— sorry, you may argue that's a denial of service threat, but we could talk about that later.

27:35Izar Tarandach But the fact that you can tamper with it, like that you have access to tamper with it. But anyway, one important thing to point out is that while I agree with you and you've got definitely a thing going on there, I think that it's perhaps it's just my the way that I approach Stripe, but I think that the categories, they apply at different layers that may go up and down the scale of, yeah, sure, this particular spoofing is fixed by this particular kind of authentication. But if you take things up many levels, you even get to business case of the thing. So all of a sudden, the threats are different. They are higher level. They are fixable perhaps by processes or something like that that do not involve some form of digital authentication or the always evolving world of certificates and whatnot. But I think that's the concept of the weakness is important, especially for those people who come from a background of no security. The simple fact that you are able to get those 6 things. Yeah, thanks. And explain to people what these 6 things are. You are already giving them so much of a leg up in terms of things to look out for at different levels, at all levels.

29:00Chris Romeo But those are not as prevalent as, or as relevant, I say even relevant, you can say they're prevalent. They're not as relevant as they were. So I guess, should we go back to the drawing board and should we create a new Stride for 2024? Because listen, I mean, Lauren, I've talked to Lauren Kornfelder about this. Everybody thinks Adam is the creator of Stride. Adam didn't create Stride. It was Lauren and another guy at Microsoft. I can't remember. I always forget who the other guy's name was. I'm sorry. But this is, they did this in the '90s, right? Or maybe the early 2000s. I mean, A lot has changed in technology. So I guess the question I pose to you is, should we go back to the drawing board? We know that the vehicle of STRIDE is powerful because it provides an on-ramp to threat modeling. Is it time to refactor STRIDE?

29:48Izar Tarandach Wait, wait, wait, two things. That paper, and now the name skips me, that paper from the '70s, '72, with the fundamentals of computer security.

29:58Chris Romeo Schrader and Salzer. Salzer and Schroeder.

30:00Izar Tarandach Salzer and Schroeder. That thing was written when Unix was like coming up at Bell Labs. Two weeks old, definitely. And it's still relevant. But I had a point about, oh yeah, yeah, yeah. So what I'm hearing you say, please correct me if I'm wrong, is that the catalog of things that some people decided to call The things we have to do to achieve security by design plug very nicely into the STRIDE buckets and therefore have fixed those categories of threats. So now we should look at a different evolution of STRIDE in order to make people look at different tools that would fix those new buckets that we have identified now.

30:52Chris Romeo That's not how I got there. But what I'm just, listen, I'm not even, I haven't made up my mind. I haven't made up my mind yet as to whether we need to— I don't know if we need to refactor this. I'm not, I'm not saying we have to. I just think, I think it needs to be extended.

31:06Matt Coles I think— wait, you're, you're, I don't, I don't agree with you that, that we don't have spoofing just because we can solve spoofing.

31:14Chris Romeo No, I said, is it, I said it wasn't relevant. So you're saying—

31:16Matt Coles so that's what you're saying is it's not relevant because it's no longer a problem, but it's still a thing. And so, and, and if you—

31:24Chris Romeo Would you agree it's a lot less prevalent?

31:27Matt Coles It doesn't matter about prevalence.

31:29Chris Romeo But why would you want to teach— this is— we're talking about a foundational piece we're going to use to teach people how to get started with this. If we give them something— hold on— if we give them something that returns no value from their perspective, then they're going to go, why the heck is Matt making me do this?

31:44Matt Coles Because you're making an assumption. You're making an assumption that technology is stagnant. Once you fix a problem, it never returns.

31:51Izar Tarandach How do I know that right now I'm talking to Chris and not to a spoofed AI-generated avatar that's trying to convince me that spoofing is not a problem anymore so that in 5 years they can come and exploit all the new spoofable stuff?

32:04Chris Romeo These are the 3 laws of robotics from Isaac Asimov.

32:09Matt Coles AI is not a robot. AI is not a robotic.

32:12Chris Romeo I think Matt's a robot.

32:14Izar Tarandach Matt's a robot.

32:15Matt Coles I'm a lawyer.

32:16Chris Romeo That's what a robot would say.

32:17Izar Tarandach I'm not a lawyer.

32:18Matt Coles I just, I just watched, I just saw the episode about, uh, on The Simpsons when Mr. Burns replaces all the workers with robots. If you haven't seen it, it's hilarious. But it dives into this exactly.

32:31Izar Tarandach But seriously, I mean, yeah, that classic problem of spoofing that we solved with that classical solution that everybody should be doing and by God, not everybody is doing. Yeah, that one got solved. But you know, we keep bringing in new people and here I go on the soapbox again. We keep not teaching people security fundamentals from the get-go. So this thing that has served us so well to teach a whole one, two, three generations of people right now, tech people, not born people. Should we deep-six it just because we sort of have a solution that's not enforced, that's not taught, that's not required, and that's not universal. Or should we just keep it open?

33:20Chris Romeo Yeah, I'm not saying dump it. So I, I, I still stand behind STRIDE as a vehicle. The question I'm asking is, should we refresh it based on current— I mean, it's been 25 years since that list was created. It's probably a lot of change in technology. We went from, we went from mainframe to personal computers back to the cloud.

33:39Matt Coles But, but by the way, mainframe, personal mainframes and personal computers still exist, right? And now you have hardware and of course software.

33:46Chris Romeo As the prevalence, as the prevalence might continue.

33:50Matt Coles Well, sure, but you don't want to create a thing that's only good for, you know, you don't want to restrict what you're looking for. Well, unless you do. I mean, if you're looking at medical devices, you probably don't need to look at things that would only affect, you know, transportation systems.

34:05Chris Romeo So how could we, how could we do that? We need an experiment. To test the relevance of, of Strag categories today.

34:13Matt Coles Oh no, he's doing it.

34:17Chris Romeo Please don't ask. He's our tech. He's, he's actually literally sending an email to Brook Schoenfeld.

34:24Matt Coles No, I think he's asking ChatGPT.

34:26Izar Tarandach To his friend Claude. To his friend Claude.

34:31Chris Romeo All right, now everybody's on the edge of their seats trying to wait for the answer, but you keep typing.

34:37Matt Coles And oh, and oh, by the way, and oh, by the way, even if it's not, you're not teaching a human, you still have to describe these things to AI so that they can recognize it. Sure.

34:48Izar Tarandach Oh boy.

34:50Chris Romeo Oh, enlighten us.

34:52Izar Tarandach Here's the question. How would you refresh the STRIDE categories to be more relevant to today's security landscape? Everybody agrees with the current question? Sure.

35:00Matt Coles Sure.

35:01Izar Tarandach I apologize, but I do not have the capability to directly update or modify the STRIDE security model.

35:06Chris Romeo I think that Claude is afraid of Adam.

35:08Izar Tarandach As an AI assistant, I do not have the authority or access to change established security frameworks.

35:15Chris Romeo I think this is a problem with prompts.

35:18Matt Coles Wait, wait, wait, wait.

35:20Izar Tarandach However, however, I can provide some general suggestions to how the STRIDE categories could potentially be refreshed to stay more relevant in the current security landscape. One, expand the spoofing category to account for emerging threats like deepfakes, credential stuffing, and other identified-based attacks. Thank you. Thank you very much.

35:39Matt Coles So wait, wait on. Stop right there. So keep spoofing and expand on what within that category?

35:46Izar Tarandach It's saying it wants to expand everything. Enhance the tampering category to address supply chain vulnerabilities, firmware/hardware manipulation, and attacks on IoT industrial control systems. This thing has been listening to Matt. Incorporate repudiation concerns around blockchain, cryptocurrencies, and the challenges of non-repudiation in distributed systems. Ta-da-da! Explore expanding information disclosure to cover data breaches, sensitive data exposure, and privacy violations in the era of big data and cloud computing.

36:19Chris Romeo No, now you're just saying— Okay, that's just a bunch of drivel. It's just a bunch of like buzzwords.

36:23Matt Coles 2020 wants their computing technology back.

36:26Chris Romeo All right, I'm running, I'm running the question through ChatGPT 01 preview, which is advanced reasoning.

36:32Matt Coles So we're on the ball here. Okay.

36:36Chris Romeo It literally says enhancing Stride right now. That's, that's the feedback and expanding threat categories. So get ready. There's going to be a big reveal here in about 20 minutes. So we get stunned.

36:47Matt Coles Yeah, this, this does not come without a cost. And by the way, and by the way, Let's talk about, let's talk about, uh, denial of service when you don't have enough power or compute to be able to run and read the line.

36:59Izar Tarandach No, that's not.

37:00Matt Coles That's not the future.

37:01Izar Tarandach No, that's not.

37:02Chris Romeo Stop denying his service.

37:03Izar Tarandach No, that's not.

37:04Chris Romeo No, I don't.

37:05Izar Tarandach Okay.

37:06Chris Romeo This actually, this actually is interesting. So the first thing that ChatGPT says is incorporate privacy threats. Stride-P. Uh-huh. Interesting. Like it would be Stripes. Yeah.

37:20Izar Tarandach Where have I learned that?

37:22Chris Romeo Address supply chain risks, include AI and machine learning threats, cloud and container security, zero trust architect. Now it's just going off the rails.

37:31Matt Coles Yeah. So enhancing STRIDE to include privacy and safety would actually, would be a good enhancement. Make it Stripes.

37:39Chris Romeo Yeah. But I like address supply chain risks doesn't work. Like you, that, that has to fit. Supply chain risk has to be in another bucket. That says expand the tampering category. Maybe that's what it is. You know what? Here's an idea. Maybe what we need is just a modern set of examples.

37:54Matt Coles We need an exploded, we need to explode out the classes of things, right?

37:59Chris Romeo I mean, that's what ChatGPT just said. Hold on. It just said enhanced tampering definition. Expand the tampering category to explicitly include supply chain attacks where adversaries compromise software components before they're integrated. That is tampering though, right?

38:13Izar Tarandach That is tampering. But then you don't have a categorization. You have a catalog and then you try to catalog everything under the sun. Well, no, it's a—

38:20Chris Romeo No, no, no, no, no. This is just a couple of examples to prove that stride is relevant today. That's the problem is everybody's saying stride is dead and nobody's fighting back and saying— Everybody's saying, well, no, it's not. So that's a terrible argument, right?

38:34Izar Tarandach Like, wait, wait, wait, wait, wait, wait, wait, wait, wait. I lost you on everybody saying that stride is dead.

38:38Chris Romeo I'm just— okay. No, I mean, so, okay, people are saying— my point is, though, that we could, we could put together a pretty quick list of like 1 or 2 or 3 examples of more modern uses of the Stride category, like supply chain compromise in the tampering category.

38:55Izar Tarandach All—

38:56Chris Romeo now I'm back on board. I'm back on Team Stride now. As soon as I saw that, I'm like, yes, that makes sense. It's relevant. It is. Those things are happening.

39:05Matt Coles It needs work. Yes. But it's still value. It still has value. And then it's, and then, and then pulling it back, right? We still haven't solved the problem of threat modeling is dead. Threat modeling is a process.

39:18Izar Tarandach Oh.

39:19Matt Coles Right? And, and we don't, we don't have time for it today, but maybe that's our next episode is, so if STRIDE is a category of threats, how do we get there? How do we identify threats? And that's a process. Right? We know the 4-question framework. We know CASTA is there. We know that there's other things that might be, might be present as methodologies where STRIDE can be applied. But STRIDE itself is not a methodology for threats.

39:46Chris Romeo I think we might need to gather our group and quickly knock out a— Assemble the Avengers. Assemble the Avengers. That's the best I can say.

39:56Matt Coles Threat modeling Avengers. Threat modeling assembles.

40:00Chris Romeo I hear the music playing right now in my head from the Avengers movies. Yes. Who's the Incredible Hulk in that situation though? Don't even answer that.

40:08Izar Tarandach Um, you really don't know how to see me pissed off.

40:11Chris Romeo Yeah, that's, that's my secret. That's my secret, Izar.

40:14Matt Coles I'm always angry.

40:15Chris Romeo All right, well, that's it. That's— we had enough fun for one day. I think we, we had some, some good debates. We had some good just good ideas unlocked there. I think Stride is alive and we just need to, we need some proof. We need to prove it. We need to provide some examples to prove that people can't just say this is dead. Well then, so supply chain attacks don't matter to you? We gotta do a little bit more work to be able to line this up.

40:46Izar Tarandach The proof is in the usefulness. As long as it's useful for somebody out there, it's not dead.

40:51Chris Romeo That's true.

40:51Izar Tarandach It's almost a Peter Pan thing. If you put it in stride.

40:54Matt Coles All models, all models are wrong. Some are useful.

40:57Izar Tarandach Exactly.

40:58Chris Romeo And with that, thanks for listening to the Security Table.

Transcript supplied by AssemblyAI. It may contain transcription errors.

Originally published as The Security Table. Part of the AI Security Table archive.