28 minSeason 2, episode 20

To SSH or Not?

With Chris Romeo, Izar Tarandach, and Matt Coles

Infrastructure, identity & resilience

In this episode of 'The Security Table,' we are back from our midsummer break to discuss OpenSSH regression vulnerability. We dig into the nuances of this race condition leading to remote code execution, explore the chain of security updates, and the role of QA in preventing such regressions.

Episode title image: To SSH or Not?

Listen to the conversation

About this episode

In this episode of 'The Security Table,' we are back from our midsummer break to discuss OpenSSH regression vulnerability. We dig into the nuances of this race condition leading to remote code execution, explore the chain of security updates, and the role of QA in preventing such regressions. We debate the necessity of SSH in modern cloud-native environments and its alternatives. Plus, we answer the critical question of who should catch these vulnerabilities first — QA teams, pentesters, or automated tools? 



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 · 28 min conversation

0:10Chris Romeo Hi. Hey folks, welcome to another episode of The Security Table. We're back after a midsummer break, kind of like Formula 1. We take a month off to retool the cars, to get the brains moving again, but We're back again to, uh, to enlighten you with, I don't know, something hopefully that's enlightening. And before we jump in too far, I want to give a quick shout out, not in a Casey Kasem top 40 kind of way, because, uh, that would be just weird. That'd be weird.

0:42Matt Coles That'd be just weird.

0:44Chris Romeo Some people, anybody under like 40 is like, what are they talking about? Google it. Look it up on Wikipedia. Now this is a, a quick shout out to Robbie and JD who are some of our avid fans. Here in the Raleigh-Durham metro area apparently like to get together and discuss the Security Table. So keep it up and, uh, give us some feedback and let us know, uh, let us know what else you want to hear us talk about. But, uh, Matt, you got an issue that you want to throw on the table here, and I have a simple answer, but I'm going to let you describe it first before I just completely wrap up the episode.

1:18Matt Coles Okay, let's— this ought to be fun. So topic for today. Regression. Regression. Sorry, couldn't resist. Anyone under 40 has no idea what I just did. Regression. So regression is a word and it means something and it is also a vulnerability and it means something. So regression, for those who don't know, I forget what the CVE is actually, it's a uh, SSH, OpenSSH vulnerability, uh, race condition leading to a remote code execution in the OpenSSH server. The important part and the reason for the fancy vulnerability name that we all, you know, naming a vulnerability is always an art and a science in and of itself, uh, regression. So regression means found and fixed and reopened. And that's what's happening here as well. So not only is this an OpenSSH vulnerability, which is particularly bad, uh, but this is a fault somewhere in the, in the supply chain that resulted in a vulnerability being opened, introduced, opened, found, closed, and then reopened and rediscovered, uh, and recapped.

2:42Chris Romeo Yeah.

2:42Matt Coles So that in and of itself is interesting, but it brings up some interesting questions about the nature of security updates, applying updates, keeping on the latest and greatest, bleeding edge versus, you know, a known working configuration. And also, you know, who, who should have caught that, right? Is that QA's job? Is that pen tester's job? This was found rediscovered by two pen testing teams. Well, actually, I should say one, folks over at Qualys Research discovered it, but at the same time, there was a university, actually, which university was, also rediscovered it. The OpenSSH vulnerability notice does have, does credit both sets of researchers for finding or rediscovering this issue. But who should have found it first, right? That's really the, The big question.

3:35Chris Romeo Well, and, and this isn't the first time that a bug has been, air quotes, fixed and then reappeared. Like, I mean, it's, it happens literally thousands and thousands of times inside of big companies where things come back around and it could be an integration problem where depending on how you're building software, You know, branches could be a branch overwrite, things like that happen all the time. Much more than gets advertised.

4:08Matt Coles Merging is really challenging sometimes.

4:11Chris Romeo Yeah, especially in non-DevOps-y kind of worlds where you're dealing in bigger changes. Like, it's one thing if you're changing 2 lines out of 2 million lines, there's unlikely you're going to get a hit, but when you're doing some type of a You're creating a new software version and you're releasing once a quarter and you have 5,000 developers all working in branches that are slowly getting merged up into trunk branches and the train, they start to consolidate in the station. There's invariably going to be some, something's going to get lost in that process and whatnot. So it does happen. We just don't, very seldom does it, is it in a magnitude vulnerability like this, which, what was the, what were the scores on this? Is this a 10.0?

4:59Matt Coles No, it's an 8.1. It has some mitigations built in and it is a timing issue. It's a race condition.

5:05Izar Tarandach It's a race condition. Race conditions are very, very hard to get all that.

5:08Chris Romeo I mean, yeah.

5:10Izar Tarandach Even reading the article, I get it.

5:12Matt Coles Yeah. On the, on the plus side, it's all there. And then if you have ASLR enabled, you have some built-in defenses. I mean, for whatever that's worth.

5:19Chris Romeo What is ASLR? I'm curious.

5:21Matt Coles Uh, address space layout randomization, for those who don't know, uh, is a means for helping when, when you have a buffer overflow to help prevent remote code execution. Uh, it randomizes the addresses of your return pointers.

5:33Chris Romeo Makes it much more difficult for an attacker to get a single exploit to work, uh, over—

5:40Matt Coles Repeatedly.

5:41Chris Romeo Yeah, repeatedly because things are always moving around. Okay. So do you want me to shut this whole thing down so we can just wrap up the episode? Let's try: The year is 2024. Why the heck does SSH still exist? Oh no, no, no, why? No, you are not going to give me a good reason in a cloud native world why I would ever need to start sshd on a machine and somehow give somebody access into it. In a cloud-native world?

6:15Izar Tarandach I'm going to tell you that perhaps it's not the best thing today, but it is definitely something that you still need running around and having access to because when everything else fails, because your IDP failed, because something in the middle failed, some— there are the answers to SSH in my experience so far. Have too many moving parts that in the middle of an incident, when you absolutely need to get to a box, that's the time when they fail.

6:50Matt Coles So, so help me understand what is, what are the alternatives? Because from what I, from what I can recall, we have VNC, Telnet, but neither of those are really good options. Big fan of Telnet and Telnet back. If not SSH, then what?

7:08Izar Tarandach There's a bunch of—

7:10Chris Romeo go ahead, go ahead.

7:10Izar Tarandach No, no, you're going.

7:11Chris Romeo Okay, cloud native, right? So we're thinking cloud native world here, right?

7:15Matt Coles You are, you are.

7:17Chris Romeo I have access to—

7:17Matt Coles I'm thinking about the scope of all Linux machines.

7:22Chris Romeo Okay, but in a, so let's, cloud native world in a modern application, you can access console services if you are still running in a console. Please don't tell me you SSH into your containers because then I will just have to, we'll have to, this will have to be called the insecurity table or the unsecurity table, right? I mean, people did that. Remember in the beginning of containers, people were like, we have to add SSHD so we can SSH into the container. Like, no, no, you don't understand. The essence of a container is if you don't need it, if it's broken, you just destroy it and make another one pop up. You don't ever try to fix it. But like in a, let's just talk containerization, orchestration, Kubernetes, containers running inside of an orchestrated environment through a, through a modern day cloud provider. If, even if I did have a box like an Amazon EC2 for some reason in my architecture, I still wouldn't SSH to it over the public internet. I would use the cloud provider's tooling. To be able to access the console of that device through a protected channel, which has been battle-tested and hardened and everything else. And so I don't see a world where I use SSH. I don't use SSH anymore.

8:38Izar Tarandach So I see two problems with that.

8:38Chris Romeo I used to, I've used it for decades, but I don't anymore because I just don't have any use cases for it.

8:43Izar Tarandach I see two problems with what you just put on the table.

8:46Chris Romeo Okay.

8:46Izar Tarandach First, many of those still use SSH under the hood. And second, battle-tested, that's a, a, a, a appeal to authority fallacy. Just because it's Amazon doesn't mean that it was battle-tested. Amazon still has its problems.

9:03Chris Romeo Of course, but it's been, but it's been used at scale by thousands and thousands of customers. And so it's been tested by attackers over and over again. And there is not an exploit available that I'm aware of today that gives me full access to an Amazon or I'm not even, we don't even have to say Amazon. It could be Google, it could be Microsoft. It doesn't matter. Large cloud provider has thousands of customers that have been running thousands of applications for thousands of years total.

9:36Izar Tarandach It's a very strong probability, but let's turn the clock back a bit. Millions and millions of people use every day Windows 95. Therefore, Windows 95 is a secure operating system.

9:52Chris Romeo Millions of people use Windows 95.

9:54Izar Tarandach No, I'm bringing the clock back a couple of decades.

9:57Chris Romeo Okay.

9:57Izar Tarandach Okay. So that's basically saying the same thing.

10:01Chris Romeo That was before Microsoft found security though.

10:05Izar Tarandach Yeah. But again, what you say is a very, very, very strong probability of being right.

10:10Chris Romeo The timeline.

10:11Izar Tarandach But I don't accept it as an absolute.

10:13Chris Romeo Okay. But the timeline. Of course, nothing, I mean, nothing is 100% secure. I'm just saying that it has been tested. And so I put more trust. Okay. I'll put, I'll, here's a, here's a premise for you. I put more trust in the large cloud providers console access than I do in SSHD today from OpenSSH, just based on track record and history.

10:40Matt Coles Well, I will highlight that. So OpenBSD, the organization, if I remember correctly, the OpenBSD was the organization that produced OpenSSH, and they have an exceedingly great track record of security vulnerabilities, especially in remote security vulnerabilities. To have this show up in SSH multiple times is, uh, is really a black mark on them or the maintainers, and I hope they get their act together. I will give you that in a cloud-native world, You're probably right that SSH is less prevalent, uh, less important, and therefore not as, as impactful. But there's a ton of systems out there, legacy systems, non-cloud native systems, edge devices, consumer devices, other things that have a need for SSH to exist so that you can do remote access, so you can do remote management. Uh, where all of those things are going to be, are potentially impacted by this and hopefully they're getting patched. Let's hope that that's happening.

11:46Chris Romeo So basically your premise is OpenSSH is important for legacy, for things that are important for anything that is not cloud native.

11:57Matt Coles Certainly. If you're not in Kubernetes and, but if you are in Kubernetes, you're using, you're using SSH to manage your Kubernetes.

12:05Izar Tarandach SSH is important as a common denominator.

12:07Matt Coles It's a secure shell. That's it. That's the goal.

12:12Chris Romeo But if it's a common denominator, then we should have invested more time and money into its overall security stance. Well, should we open— I mean, listen, do we want to open the can of worms of funding of open source projects? I started in security and system administration in the early '90s. I've used SSH since the very beginning. And it's, it has had a pathway of high, high severity, critical level remote code execution vulnerabilities. I can't remember the, I can't remember what they're all, the names of them all, but I just remember patching SSH a lot over its lifetime, over the lifetime of using it.

12:51Matt Coles And so, because they made, because they've been maintaining it very well, right? Patches are not coming out reactively. They're coming out proactively.

12:57Izar Tarandach And they are a big target because it is, I never get that word right, ubiqui—

13:02Matt Coles ubiquitous.

13:04Izar Tarandach It's everywhere. It's just everywhere.

13:06Matt Coles Yes, it is. It's in every Linux distro. And many Windows systems have SSH installed.

13:11Chris Romeo Isn't there something better? Isn't there something better that we could do that would take out—

13:17Matt Coles Is there? That's the question.

13:19Chris Romeo Yeah. I mean, but that's, if— There's gotta be something, there's gotta be a bit something better. There's gotta be a better way to do other than I get legacy, but legacy will go away eventually. Hopefully, never seems to though.

13:36Matt Coles That's the problem.

13:37Izar Tarandach Look, even if you look at the virtual KVMs and stuff like that, again, some of them use SSH. I won't say most of them, but some of them, they build their technology on top of SSH. Like they offer all the bells and whistles, but it's on top of SSH because everybody trusts SSH.

13:54Matt Coles And SSH has a ton of capability with, with, if you add PAM to it, right? If you have pluggable authentication modules, although I don't think PAM is used anymore. Maybe it's systemd.

14:04Izar Tarandach No, it is. It is.

14:06Matt Coles Uh, but you have, you have the option of certificates and passwords if you want it. One-time passwords, multi-factor authentication, RADIUS.

14:17Chris Romeo The only challenge is though, when there's a remote code in Poland. When there's a remote code execution, it never calls the PAM module or the certificate checker or anything else. It just goes right to the source and runs the code.

14:30Matt Coles Well, that's because this is a pre-auth vulnerability. This is a pre-auth vulnerability. So it happens before the authentication flow happens.

14:35Izar Tarandach So if we're going to throw away everything that has an RC, we're going to be standing with VI in our hands.

14:41Chris Romeo Awesome. I love VI.

14:45Matt Coles VI or life.

14:46Izar Tarandach That's life. No, but seriously, so, okay, so now comes the question. Vulnerabilities are a fact of life. If you are a high-profile target and SSH is as high as they come, you will be scrutinized. There's been a history both of vulnerabilities and of fixing from SSH. Most people enjoy a certain level of trust and belief in SSH. What makes SSH different from all the rest of all the other stuff that keeps having even worse vulnerabilities coming out every day? Is it because, again, as we spoke so many times, SSH is a security utility in the sense of power and water, not utility as a tool?

15:38Chris Romeo I think the challenge with it is it always leads to root access.

15:45Matt Coles No, no.

15:46Chris Romeo Well, can you, you can't run, but you can't run SSH. You can't run. Can you run SSHD as a non-privileged user? Like, could you run it in such a way that you, Matt, run a copy of SSHD that's tied to your, so you never need, uh, full system access or you never need root access. The shell.

16:06Matt Coles So you, okay.

16:07Izar Tarandach Wait, When we say SSHD, there's the official one, there are smaller versions that you can definitely run on a non-privileged port with your own UID, and those will connect only you. They are much more limited. They are not the official SSHD, so they don't come under the same scrutiny. So yes, it's possible to jail, I guess, an SSHD implementation.

16:32Matt Coles And actually with— with SELinux and other, and, and jails on, on FreeBSD, for instance, you can restrict what resources SSHD has access to.

16:41Chris Romeo Yeah. Yeah. But does anybody, has anybody successfully used SELinux? Well, in a commercial context.

16:46Izar Tarandach Oh, wait, wait, wait, wait.

16:48Matt Coles So you're opening, you're opening yet a different can of worms.

16:51Chris Romeo Listen, we've been off for a couple of weeks. Okay. I've had a lot of thoughts.

16:55Matt Coles And oh, by the way, Kubernetes, if I recall correctly, relies on SS, on, uh, SELinux and other, and other, um, Yeah, but you don't have to, you don't have to administer it.

17:03Chris Romeo That's somebody else figured out. Well, you have to finally, you do have to administer it.

17:07Matt Coles It is your responsibility.

17:08Chris Romeo No, no, no, not, not anything Kubernetes is doing with SELinux. You're not, there's no way that the administrator for Kubernetes is changing SELinux parameters. You have to have a PhD just to even consider it.

17:21Izar Tarandach No, SELinux in learning mode is one of the funniest things that you ever see going through your, uh, your system loves it.

17:30Matt Coles The point was, can nothing end up putting the system into enforcing mode and have everything stop?

17:34Izar Tarandach Oh yeah, yeah, yeah, no, I'm sure that I have had some Linux boxes in the past look at me and say, what do you want from me? Why, why, why are you doing this to me? Why? Why?

17:44Matt Coles Why?

17:44Chris Romeo That's been my experience whenever I tried to use SELinux. It's like you turn it on in that watching mode. You're like, ooh, look at all the cool things that it's going to do.

17:51Izar Tarandach Yeah, look at all the pretty permissions.

17:53Chris Romeo And you're like, oh, I can't even type on the keyboard anymore. This is cool.

17:58Izar Tarandach It's like you put it in learning mode and you let it run. And then you look at the box and say, okay, you learned it, right? So now let's enforce.

18:06Matt Coles So, so here's a perfect use case for AI. Is that what you're saying? Oh my God.

18:11Chris Romeo Figure out the SELinux policy that doesn't lock me out of the machine every time.

18:19Izar Tarandach So, okay. So we, we, we go back then to the thing that we were asking offline. Where's the QA in all this? Is it QA?

18:26Matt Coles Is QA's responsibility to find regressions? So even take it, take it out of SSH's hands for a moment. Who should be finding regressions? How, how do regressions happen?

18:38Chris Romeo What is this QA thing you talk about?

18:42Izar Tarandach When a bug daddy and a bug mommy love each other very much, then they commit And it's the wrong commit and you get a regression.

18:55Chris Romeo Uh, but no, I mean, honestly, what is the QA world look like these days? Like a lot of, a lot of people don't have QA anymore. QA has gone away in a DevOps-y kind of world where developers are supposed to be doing their own testing and automation testing is supposed to be ruling the way that we release software. And so QA seems like it is becoming a lost art.

19:26Izar Tarandach Once upon a time, every developer had that Gandalf QA person that would not let them commit. Thou shalt not commit because they kept like bringing back edge cases and bits and corners that they forgot to code for.

19:42Matt Coles There's a whole— there's a whole theory around QA testing, right? Positive cases, negative cases, boundary cases.

19:48Izar Tarandach There's a whole basket of cases and, you know, it's—

19:52Matt Coles That's what QA becomes when you end up over here.

19:55Izar Tarandach So everybody loves to hate on QA, right? And yeah, I'll hand to the sky, I used to. I'll be the first one to say so. But at some point, I I started looking around and saying, wait, there was a function around here called QA. Where have those guys gone? And when you had a good QA person, they usually were very, very good. And they knew the system inside and out, every single bit, every single skeleton.

20:26Matt Coles Where are those people? Well, I can tell you personally, as somebody who used to do QA, I joined the security team.

20:35Chris Romeo That's where they all went. I was wondering.

20:40Matt Coles Interesting.

20:42Chris Romeo No, it's, uh, it's QA is definitely an art form and I've found QA people to be some of the best threat modeling people because they have a breaking and critical thinking mentality and they also tend to do their job well have to understand the system, what you just mentioned. They have to conceptualize how everything fits together, and then they can start to come up with scenarios that sometimes developers can't see because they're looking at this little piece and the QA person was looking at the whole, everything that was around it. So, I mean, I miss QA.

21:19Izar Tarandach But, but, but they need to do a, a, a jump in the mindset there because apparently OpenSSH missed QA too. Apparently.

21:26Chris Romeo Oh, that's the beauty.

21:30Matt Coles But send hate mail too.

21:34Izar Tarandach But QA people, some of them, any of them, used to test per spec, right? Okay, this is the spec, the thing is doing what the spec says, so the thing is good. I think that those that came to security and those that are good threat modelers were those that took the next step and said, Okay, outside the spec, if we turn on the gorilla and bang the keyboard many times repeatedly on a full moon, this thing breaks. Those are the ones that I really miss. Where have those people gone?

22:08Chris Romeo All right, well, we have to reach some conclusions here. Like, what's our conclusion?

22:11Matt Coles They were replaced by a very small shell script that runs a fuzzing utility.

22:15Izar Tarandach Right? And no, and don't even get me started on, on developers and unit tests and how they like to write tests and then to code to tests and things like that.

22:28Matt Coles And no, that's not—

22:29Chris Romeo remember development?

22:30Matt Coles I was going to say, yeah. Do you remember test-driven development?

22:33Chris Romeo Still around?

22:34Izar Tarandach Oh yeah.

22:34Chris Romeo Still, it's still floating around in some languages. It's more prevalent than others.

22:38Izar Tarandach Wasn't it part of XP? Uh, oh, do I express XP in the same?

22:42Chris Romeo Yeah. Uh, Windows XP.

22:43Izar Tarandach No, no, no, no. This time is extreme programming. Well, twice the same acronym, different meanings, same episode.

22:49Chris Romeo Welcome to technology. So we got to draw some conclusions here. So, um, if I can conclude based on our discussion, we're just going to get rid of SSH.

22:59Matt Coles Perfect.

23:00Chris Romeo Good conclusion.

23:00Matt Coles And all the world comes crashing down because everyone, because everyone needs shell access. So they go to Telnet.

23:08Chris Romeo Wait, everyone needs shell access. Why does everyone need shell access?

23:12Matt Coles I'm kidding. Not everyone needs shell access, but those who do need it. When you need it, you need it.

23:19Izar Tarandach I wish everyone needed shell access. Back in those days, people knew what the system was doing. It wasn't just like a bunch of connect the dots programming APIs and libraries, and I have no idea what's going on under the hood. And if I need this thing to move faster, I'll just put bigger iron on it. I miss when people used to pay attention to what's happening in the shell.

23:47Matt Coles So we need to get rid of SSH in cloud-native environments. We need to get QA back into their desk and their job functions.

23:59Izar Tarandach Actually, question. There must be some obscure tool out there that's going to be SSH next generation and it's made of 1,000 modules written in Rust with a front in Go and something shiny on top of everything. Is there? Have you guys heard of anything like that?

24:20Matt Coles If there is, that would be awesome. Let's, uh, let's, let's be on the search, be on the lookout for it. Anyone has any suggestions?

24:26Chris Romeo I mean, I think it's an architecture solution is how we get to the next generation and it's an architecture that started back in the '90s of separating things and administering things in a controlled way versus an open way of saying, we're just going to use the internet as our, as our pathway to everything.

24:51Matt Coles Well, OpenSSH doesn't require the internet. You're opening yet another can of worms about how exposed are these systems. It doesn't.

24:59Chris Romeo What do you mean? It doesn't use IP? Like you can use something other than IP?

25:03Matt Coles It uses IP, but it does not require you to be publicly facing on the internet.

25:08Chris Romeo Sure. Yeah. Yeah. So just like any, any other service. Yeah. Yeah. No, I wasn't, but what I would—

25:12Matt Coles And what I would say is, if you show up in Shodan, uh, you've done it wrong.

25:17Chris Romeo Yeah. Yep.

25:18Izar Tarandach Totally.

25:18Chris Romeo Which, but that's the, it's, it's easy to show up in Shodan though, with SSH, the SSH default stance is people maybe that don't understand it as well as we do run it in a public way so that it's accessible on the public internet.

25:35Matt Coles But it's not just SSH then. I mean, you're running SSH, you potentially have Telnet, you potentially have ICMP because you're not running a firewall and you're not blocking on the internet.

25:45Chris Romeo Telnet doesn't exist anymore. Let's, let's, I mean, in, in the modern world, like, yeah, hopefully you don't, you don't scan for Telnet. ICMP is a debugging protocol at the end of the day, right? Like it doesn't give you any— Ping.

26:01Matt Coles It's ping.

26:02Chris Romeo Yeah, yeah, no, but I mean, it was created for debugging. It doesn't, I don't have access to some data.

26:08Matt Coles I guess what I mean is, if you have a default configuration and you're not running a firewall and your network is internet-facing and you're allowing inbound connections, you probably have more than SSH on a server exposed.

26:21Chris Romeo Yeah, I see.

26:22Matt Coles You may have, you may have etcd and you may have, uh, the Kubernetes API exposed in fact, as we've seen in the past. Yeah. So it isn't just SSH that you have to worry about at that point.

26:34Izar Tarandach Quick look at the alternatives to SSH. The only thing that's even remotely usable is that place that likes to divide everything into 4, they give us an alternative to SSH. They give you, what's it called? Privileged Access Management.

26:57Matt Coles Oh, PAM.

26:58Izar Tarandach Yeah.

26:58Matt Coles But different PAM.

26:59Izar Tarandach Oh, a different PAM. Oh, different PAM.

27:01Chris Romeo Oh, two acronyms.

27:02Matt Coles Yeah.

27:02Chris Romeo Yay.

27:03Izar Tarandach We're on a roll today.

27:06Chris Romeo Yeah. All right.

27:06Matt Coles We have to wrap it up.

27:07Izar Tarandach All right.

27:08Chris Romeo Well, we'll have to, we'll just have to, we'll have to continue to monitor the situation and and see what happens. Hopefully—

27:14Izar Tarandach So bottom line, everybody keep using SSH, but patch it.

27:18Matt Coles Patch it. Put a firewall. Don't make it internet-facing.

27:23Chris Romeo Or, and look for an alternative. And someone invented an alternative that doesn't have the same legacy challenges that exist with SSH.

27:32Izar Tarandach Interesting.

27:33Chris Romeo Well, we got to wrap it up. So once again, thank you for joining us around the security table. Shout out to our friends in Australia. Can't remember if it's Northern or Southern Australia.

27:44Izar Tarandach Treadmills.

27:45Chris Romeo Keep on treadmill. People in Raleigh, everybody else out there. Uh, we'll see you again soon for another episode of Security Tales.

27:54Matt Coles Security Tales.

27:55Izar Tarandach Dun dun dun.

Transcript supplied by AssemblyAI. It may contain transcription errors.

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