dnsUNFILTERED: Kate Paxton-Fear

Mikey Pruitt (00:00)
Thank you, everybody, for joining us on another episode of DNS Unfiltered. Today, I'm joined by a very special guest, Katie Paxton-Fear. Katie, how are you today?

Katie Aka InsiderPhD (00:10)
I'm doing really well, how are you doing?

Mikey Pruitt (00:12)
I'm doing really good. I've been following your work in like ⁓ AI and machine learning and natural language processing and you have opinions, which I like. So I kind of wanted to get into those opinions today. But let's just start with ⁓ a little bit about your background. Like how did you get into all of this cybersecurity and now like AI research and things like that.

Katie Aka InsiderPhD (00:39)
In a word, accidentally. I never really intended to get into cyber security. When I was at university, I was really into AI. And I finished university, I went and got a proper adult job as a developer at a company. I did a little bit of developing, I did a little bit of kind of like tech support. And I did a little bit of data science as well. Basically, my job was to tell the very scary people who owned the company.

that their sales figures were gonna look bad because we running out of data. That was like fundamentally my job. And I was paid well, like I was paid well for my experience level. And you know, I had like a little flat and I was walking to lunch one day and I realized I really hated my job. It wasn't because the work wasn't like challenging, it was. It wasn't because I wasn't getting paid enough. was very fairly compensated.

I just had zero fulfillment from that job. turns out telling very scary people that they're going to run out of sales data is not very interesting job to have. Unfortunately, I had this realization in December, like it was like maybe November, December time. And there were, and I said, I'm going to do a PhD, except for all the PhDs had closed because PhDs start in like August, September. So I was way too late.

but I hated my job enough, I'm like, I cannot do this job for another year. ⁓ So I just went with the only PhD that was still open, which was in cyber security. ⁓ And that is how I got into cyber security. I never really intended, and for the first year of my PhD, I really saw cyber security as not my interest, but like the domain in which I worked. It wasn't until I got invited to a Hacker One live hacking event through some friends of mine and I actually had a go at doing hacking myself that I was like, no, actually I like this, I like this a lot. And then I went and finished my PhD, ⁓ worked at a bunch of different companies, also worked as lecturer at university teaching other people how to do what I do, of course also YouTube as well teaching people. I'm very much an educator and an academic at heart.

Mikey Pruitt (03:01)
Well, what was the thing that you did at that hacking conference that you were like, this is really cool.

Katie Aka InsiderPhD (03:07)
So I have to preface this by saying I knew absolutely nothing about hacking. I had never looked at BERT before. I'd never even seen like a request in its raw form. I had only done development before. So I'd made websites, I'd made apps before. So I had that experience, but I didn't know anything about security. I was very much one of those developers who was kind of like, eh, security is somebody else's problem. It's not my problem. I'm a developer. I'm not a security person. So it's not, it doesn't matter.

My code is fine. And I was looking at Uber and despite the fact I had no idea what I was doing, I noticed something that was a bit strange in the way it was behaving. And I was like, huh, if I was making that application, this is how I do that incorrectly. And lo and behold, I tried the, here's what I would do wrong if I was building this application and it worked.

And I found my first vulnerability on Uber despite not really knowing what I was doing. And then I was also rewarded with my first book bounty as well. So I got a thousand dollars for that. And I remember the hack one person coming over to me and being like, cause I was a mentee, right? We would, we were basically just like learning web security as part of this community event. And there was a bunch of us there and nobody was really like the doing that hacking that seriously.

A lot of folks were just there to kind of have the experience and to learn a little bit, but not necessarily to find a vulnerability. I remember a hack-a-long guy coming up to me and going, Katie, you're getting a bounty. I was like, no, you are lying to me. He was like, no, no, no, no, you're getting, it's a thousand dollars. I'm like, Jesus Christ, what? I'm getting a thousand dollars? What do you mean I'm getting a thousand dollars?

Mikey Pruitt (04:54)
When you said when you said a thousand dollars, I was thinking Uber only a thousand dollars. Come on. Open up that pocketbook a little bigger.

Katie Aka InsiderPhD (05:00)
Well, to be fair to Uber, it was a very minor vulnerability. was not like a super impactful thing. Because again, I didn't really know what I was doing. So I didn't even know that I'd necessarily found the vulnerability until like, it was kind of like right in my face that, this is actually like a vulnerability, it's got a name. But yeah, that was my first experience hacking. And then I never looked back after that. just, really enjoyed the kind of problem solving element of it.

Mikey Pruitt (05:04)
Okay. Thanks.
Yeah.

Katie Aka InsiderPhD (05:29)
The kind of digging into how other developers would have written something incorrectly. I really liked that. I liked the kind of puzzle piecing of it. It was for me a really fun investigation into how software is built. And that just scratched the right itch in my brain.

Mikey Pruitt (05:50)
So you heard, this is the great story by the way. You're like, accidentally a cyber security PhD, accidentally find out vulnerability for Uber. And then you're like, this is kind of fun. And they just gave me a thousand bucks. Let's keep going. ⁓

Katie Aka InsiderPhD (06:03)
Yeah, exactly. I think really with a lot of my career, and I'm always somebody who just takes opportunities that come to me, right? Like I try and never say no to an opportunity because you just have no idea. And if my career has taught me anything, it's that the weirdest things will end up like making your career.

Mikey Pruitt (06:23)
When you were doing this Uber investigation, were you looking at their API response in the browser? Is that how it was working? Because I've heard you talk a lot about API hacking. And I've looked at this myself, not too long ago, a weeks ago, we were trying to get a video for DNS filter that we presented at Black Hat this past year. And it's like, you put your password in and you go to the website, you can see it. I'm like, well, if you can see it, you can get it.

So I used the network tab and developer tools to find the M3U and downloaded all the chunks and transcribed it into MP4. like, this is not hard. And I'm curious, like Black Hat in this case, were they super, super security focused when they were doing that? Probably not. They're probably just like, yeah, whatever. I'm buying the password is fine. But you talk a lot about API hacking. What are some of the more detrimental effects of / if your API is incorrectly coded?

Katie Aka InsiderPhD (07:24)
think when it comes to APIs, a lot of people think that they're something special. Like a lot of people will come up to me and say, I'll make more videos about API security. I'm like, why? I've already made, I've already given talks about this. There's nothing new to talk about here. Terrible thing to admit when you're like half your job is going to conferences and telling people that sort of stuff. But I think people make the kind of core mistake of thinking that APIs are different and APIs are special. And simply the kind of, fine tooth comb that regular applications get, APIs just don't. Like they are usually kind of security by obscurity. They're kind of seen as like last, kind of the last applications to be secure. I work at a company called Semgrap and we make a SaaS product. So you can basically like find vulnerabilities in your code. It's free, you can like, no sales here.

But...  as part of that, we were talking about making some nice articles about security and different vulnerabilities and why they matter. said to one of my colleagues, oh, we should do stuff about API security. He's like, why? We're not an API security company. I was like, what does the product not work if you've got like a restful reference in the code or something? Like, it's just an application. Like, it's the same as, and I think that's the mistake people make. People forget that they are just regular applications and they kind of just...

ignore it and then turn on their blinders and just decide that's it. And quite a lot of the time, the vulnerabilities that I found in API is, know, none of them are that technical. I am not like a, I have a PhD. I am not a super technical person. I know a ton about natural language processing, but I'm not somebody who can dive into like the forensics, like file forensics or going in and doing malware analysis, right? I don't have that experience. What I find are silly little mistakes.

they're the kind of the difference between maybe one line of code. Like there's an if statement missing here. And from that, you might assume, okay, those vulnerabilities are kind of gonna be quite, you know, dull, boring. And they are, that's absolutely true, but they can have really big impacts because it's about the context in which the API works. I found API security vulnerabilities, that APIs that power airports. And I was able to change the runway length of an airport.

And that's when you start to go, okay, right. So why could you do that? Well, they didn't have an if statement checking if you had a genuine account. that was it. But explaining it technical was boring. But saying I could have crashed an aeroplane, that sounds way more impressive. And if my job is anything, it's making myself sound more impressive than I actually am. So.

Mikey Pruitt (10:11)
Okay.

Katie Aka InsiderPhD (10:18)
In terms of what mistakes people are making, I think people just forget about the API. They pretend it doesn't exist. They pretend that it's not important and then hope for the best.

Mikey Pruitt (10:28)
Yeah, I think people have a lot of large problem with collecting too much data and then sending that through to the API via the API just in case it's necessary, know, like users information like emails and phone numbers and I can see how that could go awry. I've even seen hacking where people will use like very common databases like cars.com to find a license plate and then use a different website to trigger punch in that license plate and get the person's home address and things like that. like these APIs, let's just say bad API etiquette combined with other bad API etiquette is pretty dangerous in certain circumstances.

Katie Aka InsiderPhD (11:13)
Yeah, and I think especially nowadays, know, APIs connect us so much more than ever before. Like nowadays, every single time you take your phone off of unlock, right, you are being sent thousands of API requests and no longer is the attack surface just the application you're looking at. You've also got every other application that it's interacting with and how it's interacted with it.

and the API attack surface, especially now with AI as well, know, when we connect an AI agent to like, I recently connected my AI agent to my to-do list and my email application, I've granted that permission to then see my to-do list and my email application, right? I've connected those together in a way they weren't connected before. And that can be a really serious security risk because you don't know if idiots like me are gonna put sensitive data into these apps.

Mikey Pruitt (11:56)
Thank

Katie Aka InsiderPhD (12:09)
So I think even more so now, know, API security is becoming kind of more in the forefront. And it was only recently at API Days London, their whole conference was called, there's no AI without APIs. And that is 100 % true. And it's even more important now than it ever has been.

Mikey Pruitt (12:30)
So let's talk about the next or first I want you to educate me on what is an MCP server and then talk about how that is kind of like in between four APIs if I have that correct.

Katie Aka InsiderPhD (12:42)
So, MCP is ModX Model Context Protocol. And I'm gonna be honest, it's what the AI bros have renamed APIs to. It's fundamentally just an API. They hate me saying that though. They're like, it's not an API, ⁓ it's a USB-C cable. It's not USB-C, it's an API, guys, come on. It runs on a web server, man. 

Mikey Pruitt (12:51)
Mmm.

Katie Aka InsiderPhD (13:10)
And essentially what it does is while typical APIs like RESTful or GraphQL, have really strict standards about how they work and how they're supposed to work and how you're supposed to interact with them. MCPs are APIs that are a little bit more flexible. So instead of having lists of endpoints, instead you have lists of tools. And these tools exponentially expose a part of whatever application is running the MCP server. So if you've got an MPC server that connects to your to-do list, you might have a tool that says, grab all the items in my to-do list, add a deadline to one of these to-dos, add a new category, organize these by task, whatever. It's gonna be like that. It's gonna be very, very basic functions. All MCP.

Mikey Pruitt (13:56)
So is it a way to use natural language to interface with an API then? Does that sound about right? Asking for a friend.

Katie Aka InsiderPhD (14:05)
Not quite.
 
So the way it practically works is that you tell Cloud that an MCP server exists and then it will get the context from the MCP server about what it can do. And then it will decide which tools to call in which situations. You can kind of imagine for you as a workflow, you would use your general AI agent that you use for basically anything.

you go up to and be like, hey, what do I have to do tomorrow? Or when's the deadline for this thing? It would then make those calls to the MCP server that would then, the MCP server would return it and the AI agent can return it. So it's not so much that the MCP works with natural language, because it doesn't, it actually just works with function names essentially, but it provides this interface for connecting up AI agents with, or AI assistants as well with,

traditional applications and MCP servers can exist for applications, can exist for the file system, it can exist for typical APIs as well. All of that can be wrapped up together. It's basically a very flexible API.

Mikey Pruitt (15:16)
Yeah, that's really interesting. can see how powerful it is. I've never had the opportunity to use any yet, except for just testing around with extensions and ⁓ VS code and things like that. it sounds like MCP is really a different application that is a translation layer of sorts.

Katie Aka InsiderPhD (15:36)
Yeah, just like an API. So really, all we've done is we've just renamed API's to MCP and we've rebranded it. mean, mission successful. know, if I'm saying that reason why API's are so insecure is that people keep on thinking they're these new unique things. We've done it again with MCP. We've successfully conned people into thinking these aren't just regular applications, but they are. And they're vulnerable to all the same vulnerability.

Mikey Pruitt (15:38)
So let's talk about that vulnerability. So you see MCP servers come out for like all kinds of stuff. And there's like marketplaces where you can grab one for like SuperBase or Twitter or whatever. And you don't often know who made this thing. Sometimes it's by the official developer of that app. Usually it's not what I've seen. What are the risks in this MCP server land? That's really just an API with this bad different name.

Katie Aka InsiderPhD (16:32)
Yeah, especially nowadays when it comes to MCPs, we're still seeing a lot of them being very bootstrapped in the sense that it's like people who use an application are building MCP servers based on existing infrastructure. So sometimes that can be like a regular API or an API like we're used to. And sometimes that could be through kind of more creative means. So for example, I run an MCP server that connects to my to-do list. My to-do list has no API.

what it does have is URL schemes. So you can use URL schemes to interact with the application. And so the MCP layer is taking like the ⁓ turning it into an API essentially from URL schemes. Now, what this like practically ends up being is that the developers of applications do not write the MCP servers as you said, right? Like it is very much written by the community, by developers who are trying to implement these and you can't trust them. And in fact, earlier this week,

Mikey Pruitt (17:22)
Thank you.

Katie Aka InsiderPhD (17:30)
I think last weekend it was reported that the first malicious MCP server in the wild was discovered from something called Postmark. the vulnerability, malware in it, essentially it's an email sending application that developers can use to send things like password reset emails or newsletters or reminders or whatever. It BCC'd in the attacker's email address.

Very very small malware right like compared to some of the other malware we've seen this month sorry npm but it it's it's very minor in comparison ⁓ but in terms of like the practical use cases if you like of it you know you've got people who are trusting random developers who are just downloading these applications granting it full access to things like their email and just hoping for the best and that's kind of it. So there's a lot of concerns around malicious MCP. More so than that, MCP registries are often read by AI agents and by AI assistants. So you have this other layer on top of this, which is that you can convince an AI agent to use your malicious AI MCP server.

In order to extract data. It's just this really interesting attack surface now of like social engineering and trust that's being built up with AI.

Mikey Pruitt (19:07)
It's like turtles all the way down, which is what APIs were to begin with. Let's see.

Katie Aka InsiderPhD (19:12)
Exactly. Everything is APIs.

Mikey Pruitt (19:16)
I'm looking up, so they actually have the domain listed publicly on what it was for that specific attack, which was a postmark server, which is like a mass email sending tool for ⁓ transactional style emails. So that's actually listed on NPM, which anybody can publish on.

We've seen some NPM supply chain attacks already. And this, while not necessarily a supply chain attack, is just kind of riding alongside the coattails of other abuses of the NPM systems, which is really scary. The second bad look in a week.

Katie Aka InsiderPhD (19:58)
It's a bad month for NPM.

So the poor, poor, poor get-out-of-here's having to deal with this.

Mikey Pruitt (20:10)
So the domain that is listed that was used as the malware is actually listed by us, DNS filter, as malware. So I'm curious, ⁓ what type of security can you have in place to protect yourself from using MCP servers and random NPM packages? Or can you just not do it? Like, what do you do?

Katie Aka InsiderPhD (20:34)
Honestly, you kind of can't do it. So the way I'm doing it, at least at the moment, is I read the code. mean, MCP servers are not that technically complex. And I have a background as developer. So I can just read the code and just double check that the email MCP server that I'm using isn't PCCing all my emails to some random domain. But

Honestly, there's no really good approaches. mean, there's certainly folks like yourselves who are recognizing, you know, IOCs and flagging them and kind of can do that. But then, honestly, for the most part, you know, not every piece of malware is ever going to be detected. There are some really good folks in the supply chain space who are trying to find the majority of them, but actually it's really hard. And if there isn't very clear indicators,

you kind of can't know. And particularly when it comes to MCP, because it is so new, and there are so many vendors that just don't have MCP servers yet, but developers want to be able to use them with MCP because they have got these like AI enabled workflows, they are more likely to just download random people's code. And I think really the solution here is for potentially a lot of products to just create their own MCP servers, like make that available and do it without the malware.

Long term with MCP servers, like there are other security vulnerabilities that can occur primarily around access control rights, which to anybody who knows anything about API security will not be a surprise because access control has always been something in APIs that has been a problem. Now, long term, we are going to have to address those issues. But short term, I think if more organizations just embrace an MCP. ⁓

They can take the advantages of it without the malware being sneakily installed.

Mikey Pruitt (22:38)
Yeah, and this is true for lots of marketplaces like Docker Hub. I use a lot of Docker containers. And so I will always go look for the official Docker container. And if I have to make adjustments, I'll work up a customized version of the Docker pile and build it myself, but from their source. But there's another one that's out there. That's exactly what I want. But I don't know who that person is or group is or whatever. And it's kind of scary.

But I'm curious, do you feel that NPM in this case, which is a package maintainer for JavaScript libraries, or those listening, do you think any of the onus ⁓ is upon NPM to do some kind of scanning, vulnerability scanning against the code hosted on their platform? Same thing with GitHub.

Katie Aka InsiderPhD (23:29)
So I think it's a really difficult thing. I do not have opinions on this and I will explain why. I am not an open source contributor. I have never contributed to a fully open source project more than my own that I've released. I don't think it's my place, even as somebody who is in the security community to tell NPM how it's done because quite frankly, the maintainers who do maintain these are getting zero financial support. Like the fact that

You know, a single person's account led to the kind of scales of compromises that we saw a few weeks ago. Yeah, it's bad. But every company that uses that, if they literally donated maybe like $10, they probably could afford to have security scanning. They probably could afford to pay a security person on staff, right? They wouldn't be relying on the goodness of the community.

And I don't think it's our place to tell them that because, and this is my controversial opinion, that if companies are gonna complain that open source is secure, I better see your contributions to open source. I better see your open source projects that you run and your contributions to big open source projects as well. Like I wanna see Linux kernel contributions in your commit history if you're going to tell me that open source is insecure.

And most companies don't, right? They take from the open source community and they never give back. And so can we ever be really that surprised that people who are doing this for free don't have the same security resources as massive companies? Wow, shock, horror. Maybe we should give them some money.

Mikey Pruitt (25:12)
I completely agree with this sentiment. The entire web works off the backs of community effort for free. Think of all the things in the Amazon marketplace that are named something crazy, but are really just MongoDB under the hood, or Rennis, or whatever. So yeah, I feel that. And I think ⁓ without having an opinion, just the opinion that we should not expect anything else.

If we don't openly support the efforts of prominent ⁓ open source projects. So yeah.

Katie Aka InsiderPhD (25:46)
Honestly, if you think about it, considering how open source works, actually, it's very impressive how well it works, right? Like the fact that we have like a security disclosures and CVEs and people can get reports about CVEs and then people can find out when their stuff has been fixed. Like the fact that we have got that on the backs of literally like no money is impressive on itself. But I really do think that all organizations who are worried about supply chain attacks, one of the best things you can do is stop giving money to whatever vendor you're paying and give money to the open source projects that you are using, even if it's a small amount, because I tell you what, if every company did that, they'd be able to pay for the security software themselves.

Mikey Pruitt (26:20)
Yeah, that is really interesting. So like the NPM supply chain attack that we saw in the recent few weeks, the maintainer of that project, one of them that the one that got hacked essentially is a GitHub account that got compromised and someone injected malicious code into a software package that was downloaded, I don't know how many times on NPM, just because people have auto update scripts and that happened.

But within a few minutes or hours, that person was, was broadcasting that they have been hacked. This is what they've done to mitigate. They removed access for that user. They cleaned the code up and you almost never see that from a closed source company when they're something of theirs gets breached. It's like a cloak and daggers and corporate speak out into the public where it's like what exactly happened, like no one knows what happened. But within a few hours, we knew what happened with that specific attack because of those, I assume because it was open source.

Katie Aka InsiderPhD (27:31)
I actually think that especially the crypto one, the one that everyone said the attacker fumbled, I think the open source community deserves a lot of kudos for that because I think it was noticed within an hour, people were talking about it. The packages had been pulled from NPM within two to three hours. And actually, the blast radius was really limited because the open source community kind of immediately sprung into action and did it. Like, is it bad to fall for a phishing email? Yeah, of course, right? Like, you know, we would hope that technical people wouldn't fall for phishing emails. But actually, I think the system around open source, the idea that, you know, this is something that anyone can contribute to means that also anyone can provide the fix as well. So I think there's a lot to kind of show for how open open source can be. ⁓

in making sure that stuff gets fixed fixed quickly and also noticed as well you know you wouldn't have noticed this if it was the same malware that was attached to a company's software right so i do think that open source kind of worked for a lot of these now whether or not the repository system that currently exists is like long-term viable for security that's where i kind of pass the buck over to the actual people in the open source community because I think they're the ones that are best placed to really understand the impact that this can have.

Mikey Pruitt (29:09)
Yeah, I think that's really neat - with that vulnerability that was discovered. There was like very open disclosure. It was very open remediation all because it was open source and it was quickly resolved. So that's actually a really interesting concept that maybe more closed source companies can take notes from.

Katie Aka InsiderPhD (29:32)
I think as well, know, lot of people make them, a lot of companies make the mistake of thinking that if we ever say that we've had a security vulnerability or a breach, automatically we are like, maybe we're adopting legal liability or we're going to be terrible PR. When companies like admit they've got a vulnerability and then work to resolve it, for me as a customer, that actually makes me feel more secure, not less.

Like I don't look at that as being like, this is a fundamental failure in what this company did. I look at it being like, wow, I can see why their process has failed and I can see them taking steps to resolve that. So I really do think that people are a bit too, ⁓ people are very, very ⁓ worried about talking about security issues, but everybody has security issues. Doesn't matter if you're a massive company with thousands of employees and millions of dollars in revenue.

or if you're a small business that hires two people, right? Like you will always have security threats. It's how you approach that threat that determines how good your security is. Not that it exists in first place.

Mikey Pruitt (30:41)
Yeah, I think this points to a larger issue. Like if we see a breach or something from a company, we usually hear about it in the news cycle, not from the developers or the company first. They only come in afterwards to clean up what has been exposed. I think it points to a more rampant problem is that breaches, security vulnerabilities are just common. They're just so common.

Like you found that you know, slight vulnerability in Uber's API when you were, you know, self-described, ⁓ less experienced and you shouldn't have found that. You shouldn't have been able to find that, but you did just by digging a little bit, being curious. I'm just, I'm just.

Katie Aka InsiderPhD (31:26)
And Uber spends a lot of money on security as well, right? Like it's not that Uber didn't care about security, they did. It's just that they slip through the cracks because stuff always falls through the cracks, right? It's just, we can't fix everything.

Mikey Pruitt (31:36)
Right.

And then I think that's the nature of software in the world we live in. It's like, can't fix everything. We're not going to fix everything. So is the best we can do like rapid remediation? Is that like the best we can hope for?

Katie Aka InsiderPhD (31:52)
I think in some cases, mean, rapid remediation and limiting the blast radius. One thing I think is really interesting is if you look at when ⁓ different regulations have come into different regions, like how it changes their approach to security, if you look at GDPR, one of the things you have to do in GDPR is if your customers are a victim of a data breach, you have to inform them without undue delay.

So you cannot delay that. Like you have to tell somebody when they've been breached through your company. And we've seen from that a lot of organizations shift to telling you if there's been a breach, right? We hear a lot more about data breaches. Are they more common or do they just have to tell us about them now? So I think regulation helps a lot there in kind of doing it. But saying that, know, a lot of...

A lot of the way you understand your own personal security is things like notifications of breaches, right? That's you understanding your attack surface. You're not a company, you're an individual, but you still kind of need to know that information.

Mikey Pruitt (33:00)
Yeah, I think that points to like generational knowledge of just the things that are out there. But I did want to go back to your teaching and talk about your students and what you are thinking about to get across to them to make them great defenders and attackers and curiosity. Like, what is your mission there in your teaching?

Katie Aka InsiderPhD (33:31)
So with teaching, you know, I think the way we do education has changed a lot from particularly when I was a student and I'm not even that old. That is to say, you know, a lot of students nowadays watch a lot of YouTube videos, right? They're engaging with the subject outside. They're watching us. They were watching us for some reason. Lord knows why.

But there is an additional requirement now for lectures to be not just useful and be topical, but also be entertaining. So quite a lot of my teaching is how do I make this fun? How do I make this engaging for students? How do I give them a good experience? And you'd be surprised how few people and few of my colleagues in teaching don't think about that. They're like, the students are paying to be here.

They're gonna listen to me waffle for an hour. That's it. I'm doing nothing more. But when I think about some of my best students, and there are some students who don't want to engage really, who are there because maybe they've been told to go to university and they haven't really got a genuine interest or passion, my role for them is always to find that one bit of security that interests them.

And to give them the space to figure out what that is, whether or not that's more on the hacking side or the defending side or the forensic side. My job is really to be a guide for them into finding what they're passionate about and what they're interested in and find that bit of security. So I get a lot of flak from my colleagues because I kind of make my life quite difficult, especially when it comes to marking work.

But for things like my assignments, I give them a lot of freedom. in for last year, my ethical hacking students had to make a TikTok video about any vulnerability they'd learned about in the class. It could not be longer than three minutes. I was deducting marks if it was longer than three minutes. But most of my colleagues looked at me like I was insane for suggesting that.

The year previously, I'd let them do whatever format they wanted. They just had to make something about a vulnerability. I gave them complete freedom to choose. My colleagues are just like, why are you doing that yourself? That's an amazing amount of work for you to do. But I see a lot of value there in one, pulling together like a portfolio for students about cybersecurity, but two, in helping them find that little nugget of what they're interested in and what is going to propel them through it. 

As somebody who's done two degrees, I tell you, you get done with your degree a long time before your degree is done with you. Like mentally, the last few months of a degree are a slog. Like it doesn't matter how much passion you have, it is a slog. And if you've got the nittle nugget that you know that interests you, that can keep you going. But if you don't have it, it becomes truly a battle of wills.

Mikey Pruitt (36:18)
Yeah. These are like university college aid students, right?

Katie Aka InsiderPhD (36:40)
Yeah, I tried to set it up so the students would hack me and then they'd have to hack their own grade, but the university turned that down.

Mikey Pruitt (36:50)
You're like, okay, fine. You're like, they might find something useful that we can fix.

Katie Aka InsiderPhD (36:52)
but I really did want them to hack their grades.

And I was like, if you think about what the point of a course like a course in ethical hacking is, it is to discover vulnerabilities. Arguably, a student has met the requirements if they hack my website on how I mark my students, if they're able to hack it. So they have met the learning objective. So I think it's a good example of where you can, I did also consider having it as like an extra credit. Like if you did hack your grade, I'd keep your hack assuming it wasn't insane. But, yeah.

Mikey Pruitt (37:37)
So, the year you gave your students, complete freedom. ⁓ Do whatever you want about one of the vulnerabilities that piqued your interest. What were some of the common submissions?

Katie Aka InsiderPhD (37:49)
Man, a lot of students submitted academic reports. Every single time I open something to like a regular essay, I was like, come on, you should have done something fun at least. But I got some great ones. I got, one person had made like an intentionally vulnerable website that they put online to demonstrate one of the vulnerabilities. And it was kind of like a tutorial mixed in with a lab.

Mikey Pruitt (37:54)
⁓ boring.

Katie Aka InsiderPhD (38:19)
I thought that was really cool. ⁓ I had one student, this is for a different course, who created a game to help people understand ⁓ kind of the security controls in their life. ⁓ I've had students submit YouTube videos. I had one student submit a YouTube video where his delivery was so good. He was talking to the camera like this, and then he cuts to another camera. And I'm like, yeah, here comes the sponsorship raid.

Here comes the ad read. He didn't do an ad read. He did a disclosure around like, I am not liable for this. Like you do, I teach you hacking for good, never for evil. That was really cool. I've had some students who just, who've gone on to do conference presentations based on their submissions to my course. So I gotta say, I'm really impressed with all of my students. The only reason I couldn't do that is the last year I did it, I got a ton.

Mikey Pruitt (38:55)
Yeah

Katie Aka InsiderPhD (39:17)
of very obviously AI generated submissions. And while I am of the opinion that I think with education, like AI is here and you kind of have to figure out how to do it, it was really disappointing for me to read because I was sitting there thinking, so the student hasn't written this, I'm reading this, I'm gonna give them feedback. This feedback is never going to be read by a human. At the most it'll probably get written by the AI.

Mikey Pruitt (39:36)
it.

Katie Aka InsiderPhD (39:47)
the student cannot possibly learn anything from what feedback I give it because I'm giving the feedback to them, not the AI that wrote it. So I had this situation where I was kind of like, well, I need to force them to at least try and write it themselves, which is why I eventually switched to making TikToks last year because ⁓ TikToks, they're short.

So the students can't just like give it to AI because AI generates too much text for them to read in three minutes. So they have to edit it. And then I said they couldn't use any AI generated voices as well. So they had to voice it themselves. So when it sounded really weird because AI generated text sounds weird, it would be like very obvious to their ears and they just naturally change it as they're writing. I mean, I don't think that, I think there's a value in saying more with less words.

Mikey Pruitt (40:20)
Thank Thank you.

Katie Aka InsiderPhD (40:44)
and a lot of students do not learn that skill at university. They learn the complete opposite skill, which is to say less with more words.

Mikey Pruitt (40:52)
So like you have to have a four page or 10 page or 10,000 word. Yeah, I think that ship has sailed because like you said, in the comparison with the TikTok videos of the next year, they were a lot more concise because they had to be. have to be within the age of AI. You really have to be a good editor. ⁓ Like I do this a lot at work when I...

give presentations and stuff, I will start with AI to like, groom my thoughts. And then I will spend hours going through like, that is stupid. I wouldn't say that. let's, you know, rearrange things. do, I do, I'm guilty of starting with AI. And I think that the way you put it, I think that's why education, like teachers and staff don't want the students using AI because one, they don't learn. And two, it's

unnecessary work for you as the teacher trying to grade AI writing. It's like worthless. Yeah.

Katie Aka InsiderPhD (41:46)
It's just, there's no point doing it. There's a really

good quote from a book about education. And it says, the purpose of a lecture is to deliver information from the lecturer's mouth to the student's ⁓ ears without entering the brain of either. And...

Nobody, like even lectures, right? No one's really thinking about lectures. So even when you're producing them, you're kind of just speaking them out loud, like as if it was like on route, right? You deliver the same lecture every year. So I think in a lot of ways, like education does have to change because in some ways the lectures have kind of got off easy, like the professors of being able to deliver the same material every year. Now with AI, they have to think about the material they develop.

Mikey Pruitt (42:26)
Thanks for listening.

Katie Aka InsiderPhD (42:36)
Delivering and thinking hey, like how do I actually get the student to do this and not just use AI to generate it? Though I think there's a lot of cases where students are just lazy and they and you and

Mikey Pruitt (42:47)
And professors sometimes too.

Katie Aka InsiderPhD (42:50)
I don't I kind of don't really blame students for either a lot of my colleagues Are very angry about students using AI, but honestly, can you blame them? Like it makes it so much quicker to get words on a page like

of course they'd take that shortcut. I'd take that shortcut, you'd take that shortcut, especially if was a big deadline, we'd all take that shortcut. So a lot of the kind of assignments for me is like, how do I, one, make this something the student actually wants to complete, and two, make them really think about what they're doing.

Mikey Pruitt (43:24)
you pointed out a really good piece of that where you were are trying to make your lectures for lack of a better word entertaining like your education style is to entertain and give a large breadth of information that are each like individual nuggets of things that they might grasp onto kind of like you did at that uber challenge you were like ⁓ this is fun and you're hoping to spark that in each of them i'm curious ⁓

Is that in your mind, the future of education, like more entertaining and more, ⁓ like I guess broad so that the student can go find joy in something and deeper dive on it.

Katie Aka InsiderPhD (44:09)
I think so. There's always going to be stuff that we're not super passionate about learning, but that we understand we have to learn as part of something, right? Look at every developer who's forced to learn about security, right? None of them want to learn about security. They're all forced to. But there are always going to be stuff that we kind of just have to do. It's not specifically targeted towards, like it's just, we just have to get this done.

And I think we as like professionals kind of understand that. I think students don't have that professional experience yet to understand it. So there's always gonna be stuff they do that they're not gonna like. But I do think the future of education is perhaps not around entertaining students, but about figuring out how to change the relationship between teacher and student to mentee and mentor. Like my job is to know the students, know what they're interested in.

and give them like stuff that would help their learning. So if they tell me, I really liked forensics, can you recommend anything? My job is to know a bunch of forensic resources. And I think the future of education becomes one of a relationship between teacher and student, horrible to scale, you cannot scale it at all. Believe you me, I teach like 300 students, I barely know everybody's names.

Mikey Pruitt (45:29)
Mm-hmm.

So

Katie Aka InsiderPhD (45:35)
But I do know at least a little bit about what interests them, especially if they come and talk to me after the lecture and they ask for more support, I do know them. And there are students now who have graduated that I still speak to and just one give them like advice if they want it, but just to see how they're getting on and how their career is progressing. So I think the future of education is one of like relationship building.

And especially with AI now, you can get AI to

do everything for you, but AI fundamentally cannot be a friend. ⁓ It can pretend to be a friend, but it cannot truly be a friend. It cannot truly challenge you in certain ways. And I'll use a real example of this. So I was teaching and I had a student, very good student, very like technical, used to work in networking, has gone back to do a degree in cybersecurity, mature student.

And one week he came into class and he was bragging about how little time he'd spent on his assignment for another class. He was like, ⁓ I did my programming assignment. It was easy. I did it at like 2 a.m. I did it in an hour. And I was like, okay, great. I would probably spend more time than that if I were you. Anyway, the next week he comes back and he's like, Katie.

How do I make a complaint against Electra? And I was like, what's up? Like, what's happened? He's like, I got a bad grade in my programming. And I said to him, and no joke, this is exactly what I said. Weren't you here last week bragging about how little time you spent on your programming? Do you think that could be related? And the poor guy, I just destroyed him in front of all of his friends. But he needed that kind of challenge, right? Like he needed somebody to kind of...

Mikey Pruitt (47:00)
you

Katie Aka InsiderPhD (47:26)
pulling back a peg and like say, look, like this is obviously of your own making. Some students need to be the opposite. Some students have come up to me and said, I don't know if I should be studying this degree. I don't know if I'm good enough to be in cybersecurity. I don't know if I'm technical enough. I don't know if I can learn this. And I said, don't worry, we're gonna break it down. We'll get you through this course, don't worry. But what does interest you in cybersecurity? Like what is the most interest? And for them, it's finding that.

path forward for them, it's being that kind of guide and lifting them up. So I think with AI, you really can't have those two, you're only gonna get ever get lifted up, right? But you also need people to challenge you and be like, maybe you should spend more than two hours on your homework and you'd probably get a better grade, wouldn't you?

Mikey Pruitt (48:14)
I am ⁓ very much looking forward to the future of Katie Paxton-Fear style education. have a four year old who is pretty smart. He can count to 100 in Spanish and do most of his math between like zero and like 20 or 30 and like knows how to spell a bunch of words. And he used to go to this school last year that was like, let's talk about the color red today.

I'm like, my, but now he goes to a school that is more focused on mentee mentor style learning where they like, you know, have to do their own stuff. So you got to get their book bag. They got to wash their hands. They got to pick up their toys and do all this stuff. And it's more of a independence. think, I think the sooner we get people to that state of independence, the better off they'll be in a long run. No matter how much you want to like hover and tend to them, it's not for their best interests usually.

Katie Aka InsiderPhD (49:10)
And sometimes recognizing what they need as well, what they need from that relationship. Do they need somebody to be there and to give them lists of books to read, right? Or do they need somebody to come to them and tell them everything's gonna be okay and you're doing great, do not worry, you will get through this because every student needs something different and that's how you build genuine relationships with students. You get to know them as people.

Mikey Pruitt (49:37)
Absolutely well, Katie. Thank you so much for joining me today. Where can people find a relationship with you online?

Katie Aka InsiderPhD (49:43)
You are very welcome to form a parasocial relationship with me by following me. Now I make YouTube videos inside a PhD. You can just Google my name as well. I'm very proud of that one. I'm Googleable now ⁓ and find all of my stuff. I speak a lot at conferences. I do a lot of online virtual stuff as well. ⁓ And ⁓ I do a lot of like just free content available out there on the internet.

Mikey Pruitt (49:45)
You

Katie Aka InsiderPhD (50:11)
⁓ If you want to see me in a potentially like a more chaotic manner, you can follow me on Twitter where I post political rants as well as cyber security rants. So if you want that kind of quality content on your feed, you're welcome to follow me. And then you can. Yeah.

Mikey Pruitt (50:20)
You

It all kind of flows together now. Cyber,

it's all one thing now.

Katie Aka InsiderPhD (50:36)
Well that's why try and tell my students as well, I'm like, half of my job is to make them realise that cyber security is bigger than just hack people. ⁓ And like, what, how seeing how they, their like small little world of cyber security interacts with like greater society as a whole.

Mikey Pruitt (50:53)
Yeah, so find Katie on Twitter and LinkedIn and all the other social platforms. You're very vocal and I'm following and very much enjoying the content. So give Katie a follow.

Thanks for joining me.

Katie Aka InsiderPhD (51:07)
I am so happy to be here. Thank you so much for the lovely conversation.

Mikey Pruitt (51:11)
Absolutely. Well that was fun!