# Computing Has Trust Issues

URL: https://securitytable.ai/computing-has-trust-issues/
Published: 2024-07-31
Duration: 2769 seconds
Season: 2
Episode: 22
Hosts: Chris Romeo, Izar Tarandach, Matt Coles

## Show notes and chapters

Join us in this episode of The Security Table as we dive into the world of cybersecurity, starting with a nostalgic discussion about our favorite security-themed movies like 'Sneakers,' 'War Games,' and 'The Matrix.' We then shift gears to explore a critical topic in modern computing: the vulnerabilities and implementation issues of Secure Boot. Discover the intricate details of key management, human errors, and the challenges of maintaining trust in hardware and software systems. The conversation extends to the practicalities of password management, passkeys, and the broader implications of securing digital identities. 

Follow AI Security Table:

➜ Home: [https://securitytable.ai/](https://securitytable.ai/)

➜ X: [https://x.com/AISecurityTable](https://x.com/AISecurityTable)

➜ LinkedIn: [https://www.linkedin.com/company/ai-security-table/](https://www.linkedin.com/company/ai-security-table/)

➜ YouTube: [https://www.youtube.com/@AISecurityTable](https://www.youtube.com/@AISecurityTable)

## Transcript

**Chris Romeo (0:10)**: Crash and Burn, isn't that your hacker handle?

**Izar Tarandach (0:13)**: Oh my God.

**Chris Romeo (0:13)**: No, it's, it's, weren't you in a movie in the '80s or so where it was very futuristic?

**Izar Tarandach (0:23)**: It wasn't futuristic.

**Chris Romeo (0:24)**: Was it?

**Izar Tarandach (0:25)**: It was like, uh, no, it was sort of undergroundy.

**Chris Romeo (0:28)**: It was like, let's show you how the city underground actually works. All the things you don't know about. They were kind of cyberpunk-y, I think is what they were.

**Matt Coles (0:35)**: Kind of cyberpunk-y, but really not.

**Izar Tarandach (0:36)**: Kind of cyberpunk, yeah. More like cyber new wave, I think.

**Chris Romeo (0:41)**: Cyber new wave.

**Izar Tarandach (0:42)**: Yeah, there was all this neon and people in like bright, uh, lemon parachute cargo pants.

**Chris Romeo (0:49)**: There were a lot of skateboards and stuff too.

**Matt Coles (0:52)**: There were skateboards and rollerblades.

**Izar Tarandach (0:53)**: Yep.

**Chris Romeo (0:53)**: Yeah, 'cause that was kind of the, yeah, that was kind of the, yeah, 'cause you, you go to a bar and you're holding a drink and exactly what you need right now is people on skateboards flying by, right? And road plates.

**Izar Tarandach (1:04)**: What could possibly go wrong with that? Oh my God, I'm getting old.

**Chris Romeo (1:10)**: All right.

**Matt Coles (1:11)**: Try to get old. We can talk about sneakers too.

**Chris Romeo (1:13)**: So, uh, oh dude. Okay.

**Izar Tarandach (1:14)**: All right.

**Chris Romeo (1:14)**: Hold on. Hold on. Best security themed movie of all time.

**Matt Coles (1:20)**: Sneakers.

**Chris Romeo (1:20)**: You said sneakers. We've had hackers thrown on the table here. Uh, you could say The Matrix is in the category. Mm-hmm.

**Izar Tarandach (1:28)**: Before The Matrix, there was Existence.

**Chris Romeo (1:33)**: What's Existence?

**Izar Tarandach (1:34)**: He touched, ah, you see, you haven't even seen that one. He touches with, in, uh, uh, immersive gaming, VR, and, uh, has a bit of Inception in it.

**Chris Romeo (1:44)**: Okay.

**Izar Tarandach (1:45)**: Has a really good, good, uh, cast.

**Matt Coles (1:48)**: And was it based on a Neil Ste— was it based on a Neil Stevenson book by any chance?

**Izar Tarandach (1:52)**: Don't think so.

**Chris Romeo (1:53)**: Existence. Okay. We'll put that in the category. No. And it, and it's, it's written with a Z, so it's, that's why you wouldn't be able to find it. Yeah.

**Izar Tarandach (2:04)**: And it had, uh, who was it in it, man? I haven't thought of this movie. It's from 1999.

**Matt Coles (2:11)**: Okay. So just around the time of The Matrix was coming out. Yep. And, uh, which is why nobody heard of it because Matrix just, it had Jude Law in it.

**Izar Tarandach (2:21)**: It had Jennifer Lason, Jason, whatever her name is, and William Dafoe.

**Chris Romeo (2:27)**: Okay.

**Izar Tarandach (2:30)**: That's a good, a good movie.

**Chris Romeo (2:31)**: Yeah.

**Izar Tarandach (2:31)**: Spent some, uh, let me see. How long is it?

**Matt Coles (2:35)**: Doesn't say.

**Izar Tarandach (2:36)**: 97 minutes.

**Chris Romeo (2:37)**: There you go. Even better. Yeah. So. If you don't like it, you don't have to hang around for long.

**Izar Tarandach (2:43)**: But if you don't like, you're welcome.

**Chris Romeo (2:45)**: Okay. So Matt, your, your vote is Sneakers. Sneakers.

**Matt Coles (2:49)**: Yeah.

**Chris Romeo (2:49)**: Okay.

**Izar Tarandach (2:50)**: I'll have to think about this one.

**Matt Coles (2:51)**: It's a class. It's a classic. It's got everything you need. Uh, you know, drama, uh, intrigue, right?

**Chris Romeo (2:59)**: But what a cast. What a cast in Sneakers. Redford, Aykroyd, Madam President.

**Matt Coles (3:08)**: Yeah.

**Izar Tarandach (3:08)**: Yeah.

**Matt Coles (3:09)**: Ahead of his time.

**Izar Tarandach (3:10)**: What, no, what's her name?

**Matt Coles (3:13)**: What's her name?

**Izar Tarandach (3:15)**: Meta President from Galactica. She plays in Sneakers. What's her name? Yeah, I don't, I'll never remember. But anyway, my problem with Sneakers is that it reaches out to the mode of, uh, the black box that decrypts everything.

**Matt Coles (3:30)**: Oh geez, you mean like the Clipper chip?

**Chris Romeo (3:33)**: Yeah.

**Izar Tarandach (3:33)**: Yeah. Exactly. Like the Clipper Chip.

**Chris Romeo (3:37)**: You know, life, uh, follows life.

**Matt Coles (3:39)**: Life interrupts.

**Izar Tarandach (3:41)**: No. Invasion, invasion.

**Chris Romeo (3:43)**: Pop attack. All right.

**Matt Coles (3:48)**: So you're going with existence.

**Chris Romeo (3:50)**: You're going to say that's your best? No, no, no, no, no, no.

**Izar Tarandach (3:53)**: I have to think about that one.

**Chris Romeo (3:55)**: I have to think about that one. There's no time to think. This is a live show.

**Matt Coles (3:58)**: James Earl Jones was in Sneakers. Come on. You can't get any better than that.

**Chris Romeo (4:01)**: I thought it was James Bond.

**Izar Tarandach (4:02)**: I'm certified.

**Chris Romeo (4:03)**: Oh, WarGames. Come on, man.

**Izar Tarandach (4:04)**: I got a WarGames poster right behind me.

**Chris Romeo (4:06)**: That's kind of the OG, man. That is, that's, that is an OG security-themed movie. I mean, we all wanted to be the person with the high score on Galaga after WarGames.

**Izar Tarandach (4:18)**: I actually wanted to be the guy with the modem in his bedroom, but, uh, that's not cool there.

**Chris Romeo (4:22)**: You didn't have a modem in your bedroom?

**Izar Tarandach (4:24)**: Uh, uh, no. It was my bedroom in Brazil. No, it took a while for me to have a modem in my hand.

**Matt Coles (4:32)**: Yeah.

**Chris Romeo (4:34)**: I had an office. I was a weird kid.

**Matt Coles (4:37)**: Um, yes, you were.

**Izar Tarandach (4:39)**: Unbelievable. All right.

**Matt Coles (4:42)**: So as a kid, I grew up with a VT100.

**Izar Tarandach (4:44)**: So, um, you know, I had a VT100 with a modem in my room later on.

**Matt Coles (4:50)**: Green screen, green screen and clicky keyboard.

**Chris Romeo (4:52)**: And you never had eye strain?

**Izar Tarandach (4:54)**: I had a VT203 that had the orange screen.

**Chris Romeo (4:59)**: Ooh.

**Izar Tarandach (4:59)**: That was awesome.

**Chris Romeo (5:01)**: Fancy. And, uh, if you have no idea what we're talking about, just, just go to Google, put in VT100, VT203, and you'll get a picture probably from the Smithsonian Computing Museum.

**Izar Tarandach (5:14)**: Do better, download, uh, an emulator and use it as your shell. Yeah, as your terminal in whatever operating system you are in.

**Chris Romeo (5:22)**: That sounds really good.

**Izar Tarandach (5:23)**: Which should be, which should be a specific one.

**Chris Romeo (5:26)**: One that's attached to a penguin. Okay.

**Izar Tarandach (5:30)**: So what we talking about today?

**Chris Romeo (5:31)**: Today we're gonna talk about your trust issues. We're gonna have a—

**Matt Coles (5:37)**: gotta go.

**Izar Tarandach (5:38)**: Yeah. I don't know if I can talk about that with you guys.

**Chris Romeo (5:44)**: We're gonna unpack. These trust issues that, uh, now we're going to focus on trust issues. Recent trust issues in computing is the working title of this. And so, Matt, why don't you tee up the first one? You were telling us about something with Secure Boot.

**Matt Coles (6:03)**: Oh my. Yeah. So Secure Boot, uh, is, uh, as it turns out, not so secure. Well, at least some implementations of it. So, uh, for those who aren't familiar, of course, Secure Boot is the means by which you can boot a system, operating system and its firmware in a trusted way so that you can trust it to know that it's not been compromised, that it's what you intended to install. Uh, and of course then it can verify the kernel and then potentially other things as well, if it, if it so chooses, or if the system is configured. The problem is, uh, so Secure Boot runs on, uh, a key, a key pair system, right? So there's a, there's a key, uh, often there's a, a key that encrypts a key, uh, and so you can do fast verification of, of modules. And then, but there's a, there's a key that gets baked into the hardware, uh, and the, the, um, the oper— the, the system, the things that are gonna be trusted get, get of course digitally signed by that key. Or by the public, you know, the private key. And so, um, there's a set of key pairs that exist and, uh, to make development possible, of course, you have to have development keys. Now, remember, these keys are baked into hardware, so sometimes they'll bake in test keys. And that seems to be what happened here is that there's a, a small set of keys that are being used as either test keys that aren't removed or aren't properly, uh, uh, taken care of in, in the, the system, or they're, they're used as the keys and, and therefore they're, they're well known. And so that's, that allows for compromise of Secure Boot because now your trust has been broken and it's all, it's, it's, uh, untrusted turtles all the way up.

**Chris Romeo (8:01)**: Are these changeable? Are these keys, can you replace them?

**Matt Coles (8:06)**: Not often. Okay. They're burned, they're burned in. Yeah, they're burned into EEPROM or they're, they're baked into a, if you have a TPM, in theory, the TPM is, is, um, rewritable or not. You can erase and then, and then reburn or re, um, reallocate keys. Uh, is our, you okay with it?

**Chris Romeo (8:29)**: We can't hear you.

**Matt Coles (8:33)**: He doesn't know that we can hear him.

**Chris Romeo (8:36)**: We still can't hear you.

**Matt Coles (8:37)**: His audio, his audio doesn't trust us anymore.

**Chris Romeo (8:40)**: He can't hear us. We can't hear you. No, we cannot hear you.

**Matt Coles (8:45)**: We can see you and we know what you just said.

**Chris Romeo (8:47)**: It's fine. We could try to translate. So, Isar said, trusted boot. Is the best thing of all time. And he's a giant fan and he thinks no key should be replaceable anywhere.

**Matt Coles (9:05)**: Yeah. So, so devices that don't have a TPM, Secure Boot will work often by burning a key into, into EEPROM. Um, it's called a one-time programmable key. And then, uh, those keys are used to do things. Well, that information that's burned into EEPROM is often used to think, to do things like burn in the fingerprint of a certificate that is used, uh, so that you can trust the certificate because you don't have a full PKI system implementation at that point. Uh, otherwise you're using a TPM or using another trusted, you know, secure element that the hardware supplies as your, as your root, literally the root of trust. You have a hardware root of trust and then everything else is software from there or firmware.

**Chris Romeo (9:49)**: Yeah.

**Izar Tarandach (9:50)**: Can you hear me now?

**Matt Coles (9:51)**: Yes.

**Chris Romeo (9:52)**: Yes. Yes.

**Izar Tarandach (9:54)**: So yeah, no, but what I was, what I was going to ask is, uh, since I don't use a secure boot for obvious reasons, um, the TPM itself, it's part of the motherboard, right? It's not something that you can pop and put anyone in.

**Matt Coles (10:08)**: It's supposed to be, it's supposed to be, it's supposed to be tamper resistant at the very least, tamper proof ideally, although this is now we're talking FIPS territory. So, yep. For those who are unfamiliar, FIPS 140-2 originally, and I think -3 now, uh, they have different levels. There's, there's, uh, sort of, there's, uh, tamper, uh, tamper evidence, uh, meaning if it's, um, if it gets tampered with, you, you know it. It's tamper resistant, meaning it's hard to tamper with. And then tamper proof, uh, is of course, um, It cannot, cannot be tampered with, tampered with without destroying its function. Uh, so you'll see devices, you know, that have kill, kill switches or circuitry, special circuitry to, you know, if it, if you crack the case, you break the circuitry and then everything stops working. Right.

**Chris Romeo (10:59)**: So that's tamper, tamper-proofing. Most TPMs don't have that level.

**Matt Coles (11:04)**: No. Uh, HSMs, yes. HSMs usually have tamper, tamper resistance, certainly, and tamper-proof.

**Chris Romeo (11:10)**: HSM are hardware security modules for those people.

**Matt Coles (11:13)**: TPMs are. Trusted platform modules.

**Chris Romeo (11:17)**: Yeah.

**Matt Coles (11:17)**: And, uh, right. So they'll, they'll often have, they, if they're assuming they're not implemented in soft, in soft.

**Izar Tarandach (11:28)**: There is that word again, trust.

**Matt Coles (11:30)**: Don't, don't implement them in software because things will happen.

**Chris Romeo (11:34)**: Well, software is mod— you can modify it. Um, that's, you could, you could. Modify the software that is supposed to be interpreting the keys and the certificates to always return 1 when you call isTrusted? Whereas in hardware, you were supposed to be able to burn things in so that it couldn't be modified. But that does provide the downside that you've already highlighted, Matt, that you can't update it, you can't change it. So. So basically the root of this is, you could almost say the root of, no, not the root of trust.

**Matt Coles (12:13)**: It's the root of, it's, it's literally the root of trust was the wrong thing to trust.

**Chris Romeo (12:17)**: So in some of these situations they used, did they, they left, sounds like they left a test key behind and, but, but in other, in other situations it sounded like you were saying they only left the test key, like they didn't have a production key.

**Matt Coles (12:32)**: They used, yeah, they used, potentially used the, the test key as trusted key.

**Chris Romeo (12:38)**: Yeah. So this human error, is that the, the, what, what's the ultimate cause of this?

**Matt Coles (12:44)**: Uh, ultimately it has to be, I would imagine. Um, I, I don't know all the details specifically in, obviously we can't know all the details in all of the cases, but most likely this is a garbage in, garbage out problem, right? Somebody either didn't understand what they were doing when they were configuring Secure Boot. They had a script that was left in and, and, you know, because of, they may do automation for, for, um, you know, manu— secure manufacturing or for, um, uh, building these systems depending on, on, you know, how they're being provisioned. Uh, and, and so it just may not have been updated. Um, they put the wrong key in the wrong place. There's a host of reasons why that could occur. Um, but most likely it's human error.

**Izar Tarandach (13:28)**: What else can't we trust this week?

**Matt Coles (13:32)**: Well, apparently your passwords.

**Chris Romeo (13:34)**: Oh, Google. Yeah. No, no. I know all of these are passwords. They're super trustworthy.

**Izar Tarandach (13:40)**: Password 123, right?

**Chris Romeo (13:42)**: I thought you added a 4 on the end.

**Izar Tarandach (13:45)**: Oh yeah. Yeah. Because the, uh, the requirements got better now.

**Chris Romeo (13:47)**: Yeah.

**Matt Coles (13:48)**: But it's not 4 exclamation mark.

**Izar Tarandach (13:50)**: It's, it's written 4. F-O-U-R. Yeah.

**Matt Coles (13:54)**: With a, you have to put a space. You have to put a space cuz I was saying something special.

**Chris Romeo (13:57)**: That's why I couldn't log into your, and now I'm totally secure.

**Izar Tarandach (14:00)**: But no, on the, on the, on the password side of things is, uh, we keep telling people to use password managers and I still think that that's the way to go. But, uh, Google in its, uh, infinite wisdom put a password manager inside of Chrome and then they went and they broke it for a number of people with, uh, with With an update. So apparently the update was rolled in waves. 25% of Windows Chrome users already got it. And for 2% of those people, something got screwed up and their passwords disappeared. Now remember that the point of using a password manager is that you have the password manager generate the passwords and they are hopefully complex enough that it's very difficult to guess them. And that means that most people do not have a black, uh, notebook with copies of their password. So yeah, there's been a lot of, uh, forgot my password clicked on this week, I guess.

**Matt Coles (14:59)**: Yeah. This, this, we know that I, there are some sites where I only use forgot a password, forgot password, and why bother like saving one? So. If you always generate a password, as long as you have a trust, as long as the company is implementing a secure way of doing password recovery or resetting a password, that seems like not a bad idea, right? You basically get a one-time, one-time credential if you're not, if it's not being protected otherwise.

**Izar Tarandach (15:31)**: Yeah, but then actually you're what, you're farming, you're farming that security to the password of your email.

**Matt Coles (15:39)**: Well, it depends on if it's just email. Hopefully it's not SMS. I don't do that with SMS systems. That would be bad.

**Chris Romeo (15:44)**: Oh yeah, totally.

**Izar Tarandach (15:45)**: But even if you do it with, with, um, with, uh, uh, email, then you're basically just transferring the trust to the password that you have in your email. And we know that many of us keep their email open all the time. And perhaps, I don't know, if you made a text, your phone doesn't have a good, uh, lock or whatever. So it's not that we are getting more security by using the forget password, forgot password all the time. It's just that we are moving the peg one slot to the side, right?

**Chris Romeo (16:17)**: Yeah.

**Matt Coles (16:18)**: Who leaves their email open all the time? What are you, nuts?

**Chris Romeo (16:21)**: Who are you? Hey, he's always got to answer those emails quickly.

**Matt Coles (16:28)**: So you ever wonder why I'm slow to respond? It's because I don't. Keep it open.

**Chris Romeo (16:34)**: You don't, you don't let email drive your life. That's a whole other episode coming up of the, uh, you know, how to declutter your digital life.

**Izar Tarandach (16:43)**: Oh God, if you know how to, please, please.

**Chris Romeo (16:47)**: So this, this conversation of passwords made me think about, have you guys used any of these? Are they, what are they called? These things in the browsers now where you can. Effectively, it must be certificate-based. Is it a passkey?

**Izar Tarandach (17:02)**: Passkey.

**Chris Romeo (17:03)**: Is it a passkey? That's okay. So what, what, what have, what's y'all, what's your experience been with passkeys? Is that a better alternative?

**Izar Tarandach (17:15)**: You know, I, I, I have absolutely nothing against passkeys. I can safely say that I probably don't understand the whole scheme. But my gut feeling is that you're just going through your different devices and putting an egg, putting a key on each one of them and saying, this device has special value for me. And from this device, I can allow my logins by putting my fingerprint in there. And something in me mistrusts that. I, I much prefer opening my drawer, taking off, taking out the YubiKey. Putting it— did I say a vendor? Putting it in the USB slot and, and clicking on that. But hey, that's, that's just me. That's just me.

**Chris Romeo (18:05)**: Yeah.

**Matt Coles (18:05)**: Yeah. I mean, passwords, passkeys are just, I mean, so actually, Isar, can you just for our audience and I guess for me too, tell us about the theory of passkeys?

**Izar Tarandach (18:15)**: I don't think I quite can, but my understanding is that it generates a, a key pair and the private goes into your keychain, which in each device hopefully is well defended and well kept. And every time that you need—

**Matt Coles (18:33)**: protected by secure boot unit.

**Izar Tarandach (18:35)**: Oh God. And, uh, no, wait, who, who uses Windows for this thing? And, uh, damn, I said another vendor. Do we have a jar for vendors? So, Ian. Uh, yeah. So every time that you need it, you actually go and physically, uh, uh, uh, authenticate yourself to the device. And that opens the keychain and that feeds whatever needs to be fed to the browser, to the requiring application, probably some encrypted cookie or secret or whatever. And, uh, crypto magic happens and, uh, Um, yeah, you, you get authenticated to whatever it is that you were trying to.

**Chris Romeo (19:19)**: So you're at your—

**Matt Coles (19:19)**: you still have to pass a credential somewhere.

**Izar Tarandach (19:24)**: Yeah. But the credential now is a key pair that you have no knowledge of. And as we know, if you don't know it, it can't be taken from you.

**Matt Coles (19:32)**: After you can't expose it.

**Chris Romeo (19:34)**: Yeah.

**Izar Tarandach (19:34)**: You can't expose it.

**Chris Romeo (19:36)**: As Isar has been explaining passkeys, I was reading the FIDO Alliance page for how passkeys work. And, uh, did I get it too wrong? Yeah, no, you're good. Um, it's, it's, uh, it's saying that, um, during registration, the user's client device creates a new crypto key pair. It's bound to the web service domain. The device retains the private key and registers the public key with the online service. And then these cryptographic key pairs called passkeys are unique to every service. They claim, unlike passwords, passkeys are resistant to phishing. Are always strongly designed. Well, I mean, I guess it's, it's because it's taking an action to invoke it. It's not, it's not something that would be—

**Matt Coles (20:20)**: I'm laughing because this is mutual TLS.

**Izar Tarandach (20:23)**: Yep.

**Matt Coles (20:24)**: Basically, that's what they implemented. It's mutual TLS in effect.

**Izar Tarandach (20:31)**: No, but, but now you don't have like a, for the client side, you don't have an authority. You just have like, it's a, it's a shared, it's a shared key.

**Matt Coles (20:37)**: It's a shared key pair, but it's a pre, it's a pre-shared key pair for mutual TLS.

**Chris Romeo (20:43)**: But then what they do is they, they tie it to some biometric.

**Matt Coles (20:48)**: Yeah. Yeah.

**Chris Romeo (20:50)**: There's some, so it's multi, so it does give you, it's multifactor-ish.

**Matt Coles (20:55)**: Multifactor-ish. Yeah. It's a key pair that then gets unlocked. You have a key encryption key basically, or probably that You know, is used and you have multifactor to unlock that key.

**Chris Romeo (21:06)**: Yeah. Or a FIDO security key as well can unlock it.

**Matt Coles (21:09)**: Yeah. Because a FIDO key or, or a, uh, you know, the other key that is mentioned or Titan key, you know, they all just hold credentials, right? They'll hold either a certificate, a password of some sort, or, uh, metrics data to be able to do, um, a, um, a pattern match. Yeah.

**Chris Romeo (21:30)**: All right. So we basically trust passkeys is what you guys are telling me.

**Matt Coles (21:37)**: Well, they can still be compromised. I mean, uh, if you, if you have access to the hardware device, uh, that's, that's a little bit harder, but if you have access to the software device, you know, bits, uh, bits can be— an interesting question would be, if, if somebody were to dump memory and get bits of keys, first, would you know? And then second, uh, how easy is it to reconstitute the information? Um, so I wonder, has anyone, if anyone's ever done any analysis on, uh, on passkeys, if you have access to memory.

**Chris Romeo (22:16)**: Yeah.

**Izar Tarandach (22:17)**: But wait, is, isn't that, uh, Isn't that the same as doing an analysis on any use of, uh, uh, crypto material that lives in an enclave? Because I'm guessing that depending on the hardware, if it's capable to do enclave crypto operations, then those keys are not going to appear in main memory.

**Chris Romeo (22:37)**: They shouldn't be, or maybe if they are, it's only for a blip. It's not something that sits in memory because, right, the Passkey is an authentication mechanism. Once the authentication transaction has been successful, they better dump any of that stuff and, and zero out any of the memory that was used to hold those things. I mean, that's coming back around to FIPS, right? That's a FIPS principle as well of cleaning up crypto material wherever you use it. You don't leave, you don't leave remnants of it sitting around.

**Matt Coles (23:11)**: I mean, a good implementation would be, you know, Chrome or or, you know, Bing or whatever you're using, uh, not Bing, uh, Edge, um, to, uh, call to the OS to do the functions, right? So there's no, there's no crypto being done by the browser itself. It should all be done by, by the OS or by, uh, kernelmod or something.

**Chris Romeo (23:31)**: Yeah.

**Izar Tarandach (23:31)**: But then we go back to the original question. So we have passkeys and it's great and it's whatnot, but then you get a patch on your browser and your, and that patch just says erase all passkeys that are in there, like just erase them. It's not what's used, what happened to Chrome. Chrome was more of the password manager function in the browser itself.

**Chris Romeo (23:52)**: Well, I think you need to use different components though. I think part of trust is the age-old wisdom, don't put all your eggs in one basket. And I don't really know why is it if I drop it because all the eggs will break? I don't know. But if you layer your trust using a single source, then it's tough to argue that you have a trustworthy solution if you're using components from the same provider, the same software house, the same vendor.

**Izar Tarandach (24:30)**: But you, you— there is a—

**Matt Coles (24:32)**: go ahead, sorry.

**Izar Tarandach (24:33)**: No, you, you go, man.

**Matt Coles (24:35)**: No, you go.

**Izar Tarandach (24:36)**: The, the, you used a word in there that I really liked, the trustworthy. Because when, when Microsoft went ahead with their Trusted Computing initiative, it was trusted. So like, we're going to write these things and you people can trust it. But nobody that I know came out with a trustworthy initiative to say, we wrote this, we designed this in a way that you can trust it. And here's the proof that you can trust it. We're just told to, hey, you know what, trust it, you know, because we say so.

**Chris Romeo (25:08)**: I, I, I had this, uh, yeah, it was a conclusion. It was a conclusion.

**Izar Tarandach (25:13)**: I had this teacher back in high school that, uh, every time that we, he, he was a physics teacher. And every time that we would ask him something beyond the, the limit of high school physics, he would look at us and say, well, because Newton said so. And I feel like we're back in the same place. Well, because Microsoft said so, right? It's trustworthy because somebody said so.

**Chris Romeo (25:36)**: Yeah.

**Izar Tarandach (25:38)**: And people still don't have the, they are still not naive, but they are not informed enough to come and ask, hey, prove to me that you are trustworthy.

**Chris Romeo (25:47)**: Yeah. And this is, I mean, my, my conclusion here is I would never use a password manager that was part of a browser. No, because I'm putting, I'm taking two disparate pieces of data or one very important piece of data and I'm mingling it in the realm of, in the control of one player. And so I want multiple players and that's why I use a password manager, for example, that doesn't store, they don't store passwords. They're not in the business of storing passwords. So they give you X number of storage providers that they've integrated with that you can then use to store your passwords, all of the popular ones that you would think of.

**Izar Tarandach (26:37)**: I wish I trusted myself enough to do that.

**Chris Romeo (26:40)**: What do you mean?

**Izar Tarandach (26:44)**: I'm sure that at some point I'm going to bark things and erase the wrong file. And lose all my passwords in the process. Yeah.

**Chris Romeo (26:51)**: Yeah. So it's, so my, that's why you have to back it up. Yeah. No, I have a, so the, the solution that I use, you, it, it, um, uses the storage, but it also has a local, it also does a local backup copy. So, ah, I, I'm never, I'm never in, but also on those storage providers, you have backup automatically, right? Like who stores files on their local computer anymore? Like we have these cloud services that are integrated. You guys do 'cause you're security nerds. But you can also lose your data pretty easily when you have a hard drive. Like, oh, hard disks still fail. Oh man, I didn't know that was going to happen.

**Izar Tarandach (27:27)**: SSDs still fail.

**Chris Romeo (27:28)**: I know, that's what I'm saying. That's why, that's why we have this cloud thing, this, uh, this multi-cloud world.

**Izar Tarandach (27:37)**: Um, yeah, I think, yeah, you think?

**Chris Romeo (27:39)**: No, no, I'm just, I mean, this, this is part of unpacking. Trust is like, what are the, what are the core principles that define trust? And it's a certain amount of paranoia, I think is one of them.

**Matt Coles (27:58)**: So it's, I mean, it, trust is all about, uh, uh, having a clear provable identity. Of something, right? That doesn't have to be a person. Uh, and a chain of evidence that you can verify.

**Izar Tarandach (28:19)**: And that's trust.

**Matt Coles (28:20)**: I mean, that is trust, right? That, that you have— Nice. Dog in the background. Uh, so we do, you do have a notion of something that you can say is that thing. And then that it has certain properties that you can verify over time, right? That is the concept of trust. And so that's what our challenge is when it comes to security, right? Is how do we prove identity and maintain that identity proof over time? And that, and sorry, and its integrity over time, right?

**Izar Tarandach (28:57)**: Yeah. So just to go back into some, themes that we keep going back to. So we have trust in all its forms. We have the reasonable security, meaning you should do as much security as you need, no more, no less. We have, uh, uh, what was it, what, what was it that we, we threw in there? Uh, secure by design and secure by default. And if we know all that and we put all that together, why is it that, what's missing here so that we can actually secure the world? And I go back to the care thing. I still feel that with all these breaches, all these problems, all these things, and I experienced this whole week. When I met people, like normal people who know what I do for a living, and they immediately would ask, hey, hey, what's happened? Like, what's this whole thing that no Windows works and all that? Went to the gym, half of the monitors in the gym don't work for the same reason. And people are like, uh, oh wow, how could you guys do this? I'm like, okay, first of all, it's not me, but, uh, second, like What did you expect? And they're like, well, we just expect things to work. That's the thing. People still don't care. And for example, this conversation about the password manager and the ways that Chris is managing all that stuff. Can you really? I can count in the one finger of one hand the amount of non-technical people that I know that would be willing to go that way and take all those steps and manage it by themselves. Yeah. And care enough about it. And this one person actually has a bunker in their, in their backyard filled with cans.

**Chris Romeo (30:57)**: So you're not supposed to tell anybody about my bunker. Oh, you were talking about someone else's bunker.

**Izar Tarandach (31:02)**: That's me. No, no, don't back me up.

**Chris Romeo (31:03)**: Yeah. But no, I mean, to, to address, to, to, to answer your question. The reason we can't secure the world is really about discipline and convenience. Convenience. Yes. On the user, on the user experience side. Yeah.

**Matt Coles (31:23)**: Easy.

**Chris Romeo (31:24)**: The easy button. But it's really a discipline problem. The reason why, if everybody would just do what we prescribe from a security perspective, threat model every story and use, so, so you're implementing secure by design, secure by default, uh, use checks and balances within your secure development lifecycle and do this universally with every release that you have. We could, we could get 90% better than we are today, but the problem is people run businesses. They don't run security shops where security drives everything. Because if you get to that level where security drives everything, it's a world that the three of us would love. We'd be like, this is awesome. We love it. Of course we have nothing to do.

**Matt Coles (32:19)**: So we have to take up playing golf, which this will be, go and do other things.

**Chris Romeo (32:23)**: I mean, security table golfing show is what this would become, but it's the lack of discipline.

**Matt Coles (32:30)**: We would all have jobs in those companies that were building cool things with proper due diligence and care.

**Izar Tarandach (32:37)**: Yeah.

**Chris Romeo (32:38)**: I don't know.

**Matt Coles (32:41)**: I would, I'd be playing with robots, I'm sure, or something.

**Izar Tarandach (32:44)**: Oh, totally.

**Matt Coles (32:45)**: Or threat modeling.

**Izar Tarandach (32:47)**: What's, what's the driver that we are missing here to get there? Because we, we, we keep playing Cassandra. We keep telling people these bad things are going to happen and nobody believes us.

**Matt Coles (32:59)**: I, I don't have, I don't have it handy, but it's, it's, um, it's, uh, it's this, it's this green paper, paper thing that, uh, people exchange for stuff. Right. Uh, but that's no, but that's, that's not, you want some, we gotta give you something. No, I understand you.

**Izar Tarandach (33:18)**: I, I, I get what you say, Matt, but you know, it's, it's like, we're not asking people to do something unrealistic here. Perhaps we are asking them, okay, slow down 1% so that we can run later. It's, it's not infeasible. It's not—

**Chris Romeo (33:37)**: the challenge though is that we're all sitting on the same side of the table and we're saying this is only a 1% impact, but is it really a 1% impact? When we think about the depth of engineering. And I think that's, I think we're uncovering a bit of a challenge that we can have as security people in that we tend to see the world with security-colored glasses. And we also think things are a lot easier than they actually are because they're easy in our mind. Because why don't you just run SAST on every, in every build pipeline? Or every time a developer checks in code, they should just run a SAST tool. And they should just fix any of the problems that the tool identifies. That's easy for us as security people to prescribe, but we, we gloss over false positives. And when the tool, they run the tool and it generates 5,000 results, like we miss out on all of those things that the developer experience side of the house sees as blockers, stoppers, or things that would keep them from using this discipline that we're asking them to use all the time, unilaterally.

**Izar Tarandach (34:53)**: I get what you're saying. The thing is that I always try to bring together my security practice to actual developers. And over the years, the, the, and I can hear my voice saying this in 2010, 2012. We, we've been saying that you have a choice to do your preventive care and see the doctor once a year, or you can wait until later and run to the ER with things oozing out of you, right? And over the years, to me, that's changed into, from, from like, from medical care, the, the, the analogy changed into dental care. You can brush and spend some time flossing every day, or you can see as the cavities and the gingivitis mount up, and at some point you can't eat anymore. And we, we are at the point right now that people are looking around and saying, security people, what do you want from us? We can't eat anymore because you are pointing at too many cavities in our teeth. So it's, it's a self-created problem. If, if you don't do the, the, the basic hygiene in the beginning, if you don't adopt a mindset that I have these number of tasks that I have to do so that I don't have to suffer and pay the time and pay the money. Yeah. In the end.

**Chris Romeo (36:20)**: Yeah. But that, that, uh, that train has left the station.

**Izar Tarandach (36:24)**: It hasn't. That's the thing.

**Chris Romeo (36:25)**: A long time ago.

**Matt Coles (36:26)**: No, I also wanna, I also wanna throw out there. I, there's something else we need to throw out in this conversation. The problem is, the problem is the analogy is bad, right? It's because dentistry operates on a set of teeth and teeth are teeth, right? There's one teeth, one set of, one type of teeth. There's one technology of teeth that they're dealing with. And so all of the dentistry works on that technology.

**Chris Romeo (36:51)**: Single stack. Single stack.

**Izar Tarandach (36:53)**: So we clearly don't watch the same movies. Yeah. Okay.

**Matt Coles (36:57)**: So, so the problem is the problem, a problem. It is not the problem and it is, it, but it is related that because there's so many, how many programming languages are there?

**Chris Romeo (37:13)**: 1,217.

**Matt Coles (37:14)**: How many of those programming language and how many operating systems are there and how many of those programming languages and toolkits and everything else there's this just magnitude of problem, right? All it takes is one of those, let's take a X, X was XZ Utils, right? All it takes is one piece to not have the due diligence to cause a ripple effect throughout. And so when, when you're talking about you're generating tons and tons of issues, well, there's tons and tons of, tons of security weaknesses. Right? The, the CWE database is full of possible flaws for security. Oh, oh, oh, we're not talking privacy.

**Chris Romeo (37:56)**: No, no, no, no.

**Matt Coles (37:56)**: We're not talking other things.

**Izar Tarandach (37:58)**: Wait, wait, wait, wait, wait, wait. I'm going to push back on the analogy and it just proved my point.

**Matt Coles (38:02)**: Oh, fair.

**Izar Tarandach (38:04)**: There is one set of teeth, but you have different kinds of teeth. You have incisors, you have canines, you have molars. Each one of them have a different geometry, but the fundamentals are that you have to brush and floss. If you brush and floss all of them, doesn't matter which kind of tooth it is, good things are going to happen. Now look at the CWE. The CWE is a tree. It has a root, it has medium leaves, and it has all the branches down to the minimum detail of a CWE. If you adopt the fundamentals right, if you do your hygiene right, you're going to be killing those, uh, those, uh, weaknesses. At the higher level of the tree. So you won't have to worry about the branches and the leaves on the bottom. So the analogy stands.

**Matt Coles (38:51)**: I sit corrected.

**Chris Romeo (38:54)**: For the first time on the Security Table, somebody has admitted that they accepted.

**Izar Tarandach (39:00)**: No, no, no, no. I admit all the time when I'm wrong. I may not say that somebody is right, but I admit that I'm wrong. Uh, so you see, flossing and brush, that's, that's my whole point. I gotta run back to fundamentals.

**Chris Romeo (39:18)**: I gotta go floss right now. Yes.

**Izar Tarandach (39:21)**: Ooh, I knew I forgot something this morning.

**Matt Coles (39:23)**: Yeah.

**Chris Romeo (39:24)**: Uh, no, I mean, it's, this is, this is a challenging issue. There is no single solution for, because there's certain there are multiple trillion variables that would have in this equation.

**Izar Tarandach (39:42)**: But Chris, you're like one of the most successful cases of, uh, uh, actually teaching secure development and secure design to developers on a day-to-day basis, not perhaps as an instructor, but as a creator, as a, everything that you put together and everything that you successfully sold out there. Tell us why it's so hard to educate developers.

**Chris Romeo (40:07)**: I mean, it's ultimately about time is always the challenge is, is I found towards the end of, of my, uh, days at Security Journey, didn't find anybody who pushed back to say, oh no, our developers already know this. Nobody said that. Or, and nobody said our developers don't need this. The challenge comes down to why doesn't everybody do it? Well, because they're under-resourced already to begin with. And this is another thing on top and we have to trade away these features to be able to invest in this foundational thing. And it's a, it, yeah, as Matt's making the, the dollars and ultimately people run businesses and businesses have bottom lines and businesses have revenue targets and things. And that's—

**Matt Coles (41:06)**: and by the way, and by the way, we have open source projects that don't have necessarily those factors, but have a different problem of not enough resources, not enough time, somebody who's volunteering their time to do work. And that's why we can't solve this problem with open source either.

**Chris Romeo (41:23)**: Yeah, because they're focused on building a feature because it's, that's what people want. That's what they're, that's what they think is cool. It'd be great if they thought threat modeling their open source library was cool or flashy or was going to get a press release, but no one's going to write a press release about the developer who did a threat model of an open, their open source library pieces.

**Matt Coles (41:46)**: Actually, that would be an interesting story to tell. Perhaps we should open source developer that had an open source developer that, that knew enough about security and wanted to protect its users.

**Izar Tarandach (41:58)**: Or perhaps they didn't even know much about security, but wanted to at least do the right thing and threat model it.

**Matt Coles (42:03)**: Yeah.

**Izar Tarandach (42:03)**: That would be a really cool story to tell.

**Chris Romeo (42:04)**: I'd, I'd love to hear that story if we could find somebody who was in that category. Huh.

**Izar Tarandach (42:12)**: Okay. Challenge accepted.

**Matt Coles (42:15)**: Challenge accepted. That sounds like a future episode waiting to happen.

**Chris Romeo (42:19)**: Yeah. Somebody's out there.

**Izar Tarandach (42:20)**: Let's go hunt for someone out there that did this. There must be someone.

**Matt Coles (42:26)**: I'm sure there's someone. How many projects are there on GitHub?

**Chris Romeo (42:32)**: Yeah.

**Izar Tarandach (42:32)**: Speaking of which, GitHub is another one that we lost trust this week. Because they figured out that deleted stuff on GitHub is actually accessible. And what's one of the things that gets most deleted from GitHub?

**Matt Coles (42:46)**: Credentials.

**Izar Tarandach (42:47)**: Ba-da!

**Matt Coles (42:48)**: Hard-coded passwords.

**Izar Tarandach (42:50)**: Right? So yeah, well—

**Matt Coles (42:53)**: So Git Purge doesn't work anymore?

**Izar Tarandach (42:55)**: I didn't quite read the whole story. Just so much stuff happened this week. But let me see if I can give it a quick— no, I'm not using ChatGPT. I'm using the register.

**Chris Romeo (43:07)**: Which is usually, it was written by ChatGPT.

**Izar Tarandach (43:11)**: Oh, you know, sometimes you feel like, but you always have the, uh, BOFH.

**Chris Romeo (43:18)**: Yeah. Page to save you. I don't know. They went, they went so commercial. I used to love that site.

**Izar Tarandach (43:24)**: Oh man.

**Chris Romeo (43:24)**: I used to love their snarkiness that was in every article that they wrote.

**Izar Tarandach (43:30)**: I used to totally identify with it. But yeah, you're right. Lately they, they became like way too—

**Chris Romeo (43:40)**: they went corporate. Everybody does. But coming back around, it's about the money. You're running a business.

**Izar Tarandach (43:44)**: Here you go. Here you go. Data from deleted GitHub repos may not actually be deleted. Um, research at whatever have found or arguably rediscovered that data from deleted GitHub repositories, public or private. And from deleted copies, forks of repositories isn't necessarily deleted. There's a cross-fork object reference. Oh, that's a new term. It even got a new term. Okay, cool. Yeah.

**Chris Romeo (44:16)**: I mean, anybody can add a new term. There's no committee. It's too abrasive.

**Izar Tarandach (44:19)**: Oh, wait, wait, wait, wait. Yeah. As we all know, a C4 vulnerability occurs when one repository fork can access sensitive data from another fork. Including data from private and deleted forks. Interesting. The firm showed how one can fork the repository, commit data to it, delete the fork, and then access the supposedly deleted commit data via the original repository. It's almost like when you used to delete a, uh, a DOS file and then go and change the, the, the path and then delete it again.

**Matt Coles (44:51)**: Let's talk about COVID channels and ADFS and all that sort of stuff.

**Izar Tarandach (44:57)**: Mm-hmm. Wow. Wow, people. How, how come this didn't become a major thing?

**Chris Romeo (45:04)**: So there is no trust is what I learned today. Trust is, uh, that's the only thing that I trust is that there is no trust. So that's, that's a good conclusion. That's a good place to end. Good place to end this, this conversation. So, uh, I don't know, we went somewhere and then we turned a couple of corners, then we spun around the block. And then we landed somewhere. And this is a challenging issue. It's just not, this is, it's not something that's, of course, do these 3 things, even though Matt laid it out well, as far as what trust is, it's just really hard in implementation. It's easy on the whiteboard. You can probably use 20 words to describe it and say, that's what it is. But living it, breathing it, and implementing it in everything that you build and everything that you sell.

**Matt Coles (45:49)**: Funding it.

**Izar Tarandach (45:50)**: Funding it.

**Chris Romeo (45:51)**: That turns out to be a challenging problem. So folks, thanks for listening to another episode of the Security Tip.
