Anmol: Welcome to the AI Security Update, where we discuss the latest in AI and security. Today, I'm so excited to introduce our guest for this episode, Lisa Ventura. She's a very famous cybersecurity leader and expert based in the UK. So she is an author. ⁓ wrote several books, including Rise of the Cyber Women. ⁓ also the founder of Cybersecurity Unity. ⁓ and she was recognized as a member of the Order of the British Empire in 2023, which is a very big accomplishment in the UK, Lisa, thank you so much for being on the show.
Lisa Ventura MBE FCIIS: It's absolutely wonderful to be here today. Thank you for having me.
Anmol: Lisa, can you introduce yourself a little bit more on your background and your experience related to AI and cybersecurity?
Lisa Ventura MBE FCIIS: Absolutely. So as you rightly mentioned, I've published books, I founded Cybersecurity Unity some years ago, but more recently, I've actually founded what I believe to be the world's first membership body and trade association that's solely dedicated to the intersection of AI and cybersecurity. and that's called the AI and cybersecurity association, through which we'll lobby government for the safe, secure, responsible and ethical use of AI, raise awareness of the use of AI. ⁓ I guess not so much for the huge conglomerates, the big businesses, ⁓ but more for the smaller businesses that may be using it and especially using a lot of the learning language models day to day. but not sort of really thinking about the security ramifications around those. So that's very much ⁓ where my work is going to be focused on going forward. And I've also had another book published with Kogan Page, which is called Artificial Intelligence in Cybersecurity. And that covers how AI is transforming the threat landscape and how we defend against it. And it was during that process of writing the book and researching the book that I had the idea for the trade association. ⁓ Everything I do as well around the cybersecurity awareness training space, it's always empowerment led, trauma informed, and importantly for me, neuro inclusive, because I'm also openly neurodivergent. And I was diagnosed with autism and ADHD a few years back.
Anmol: Thank you so much, Lisa. We also had Nathan Chung on our show, who's also openly neurodivergent and he's a neurodiversity champion. And I think that brings a unique angle to how approach problems and solve problems. So ⁓ for cybersecurity, we need all skill sets. ⁓ there anything else you want to add about neurodiversity or ⁓
Lisa Ventura MBE FCIIS: I mean, me, I've done a lot of work ⁓ research that area and also act as a mentor and support people that are neurodivergent to get into careers in ⁓ cybersecurity. many studies have shown that those of us who are neurodivergent are actually really well suited to careers ⁓ within cybersecurity space, ⁓ mainly because a lot of the time we have strong pattern recognition, have strong hyper focus and so on. And that lends itself really well to many different roles within the cybersecurity space.
Anmol: Yeah, and I would say that also applies to AI security as well. When you look at the different attacks, one of the classic detections is anomaly detection, right? So seeing what doesn't match the pattern. And that, again, is pattern recognition, which is very important.
Lisa Ventura MBE FCIIS: sure. Mm-hmm.
Anmol: based on your experience, what challenges have you seen related to AI security? ⁓ You mentioned small businesses, but what are the challenges particular?
Lisa Ventura MBE FCIIS: I've ⁓ quite a few and that led me to the founding of the AI and Cybersecurity Association with my research for my recent book. ⁓ I think for me, it's the huge rush ⁓ I see organizations have to adopt this technology. And a lot of the time I'll sit down with them and they'll say to me, we need to roll out copilot or Gemini or whatever tool they've chosen. ⁓ And I'll say to them, but what specific business case do you have for the adoption of this technology? And what are you looking for AI to solve in your business? And they'll turn around and say, well, we don't know, but everybody's got it. So we need to adopt it. We need to have it. And there is this huge rush to adopt this technology. ⁓ And for me, I do a lot of awareness raising to businesses to say, It's great that you want to adopt this, but make sure you have those guardrails in place. Make sure you have those AI policies in place. Make sure as well you adopt and use the tools that will be the most benefit to your business and your organization. And really importantly as well, have mechanisms in place ⁓ to have approved tools for use within your organization. and also be aware of any shadow AI, is becoming a bigger problem in shadow IT, I think at the moment. And be aware of that and that people might be using different tools almost sort of underground. I mean, I've seen organizations that have just gone a blanket ban, no, we're not using any of it. And that's accelerated their shadow AI. So... For me, those are certainly some of the challenges that I'm seeing in the space. Also from an AI and security perspective, a lot of people that are involved in the cyber awareness and training arena, it's rendering a lot of the traditional advice that we would often give around spotting what might potentially be phishing email simulations and so on. AI deep fake videos, they're getting so, so much better and so much more convincing. And a lot of that traditional advice of look at the spelling, look at the grammar, look at how an email is constructed and then so on. With attackers using these models to produce these emails, you the average person on the street will look at it and potentially think that it's a genuine ⁓ email when in fact it's a phishing attempt. So those are certainly some of the biggest challenges that I'm seeing in the space at the moment.
Anmol: you need to think the use case and what your business actually does and analyze and see what data do you have? do you actually need AI to process the data? I always say sometimes a manual approach is better than AI if you can use a traditional spreadsheet. Let's say you have like 10 data points. You don't need AI for that. AI is very good for like thousands and thousands of data points, a lot of data points, you're trying to find one needle in a haystack ⁓ that's not something a person can do, then maybe try to use AI. But it requires a lot of analysis and determining whether or not you really need to use AI for your use case or for your business. So definitely agreed there. And Lisa, we see AI and AI security in the news. People hear about it almost every day, and there's always a new development. based on what you've seen in the news, whether it's global news or local news to you, what a current event that you think our audience should know about?
Lisa Ventura MBE FCIIS: Absolutely. So for me, there was quite a significant one that broke ⁓ a couple of weeks ago, and that was actually the Merkle ⁓ So ⁓ is actually an AI recruiting startup, and it was hit with a supply chain cyber attack, ⁓ and it involved ⁓ open source project Lite LLM. ⁓ for anyone that's not ⁓ with this, Merkle recruits domain experts such as scientists, doctors and lawyers. ⁓ And they also then provide training data for major AI companies and their customers ⁓ actually include Anthropic, OpenAI ⁓ and Meta as well. ⁓ for me with this, we're actually talking about a company that sits right at the heart of the AI ecosystem. And what happened with this particular breach is that a hacking group called Team PCP, it actually compromised Light Allalem. And that's actually a widely used open source library for connecting applications to AI services. ⁓ Merkle was actually one of lots of companies that were affected by this. ⁓ then separately, another extortion group claimed responsibility and shared samples ⁓ stolen data, which reportedly included information from Slack, support ticket data, and even recordings of conversations between Merkle's AI systems and contractors. And for me, this really matters a lot because according to unconfirmed reports, data sets used by some of Merkle's customers and information, especially around some quite secretive ⁓ AI projects. may have been compromised. So the implications for this are huge because you compromise one link in the supply chain and that could potentially then grant access to the training data and project details of some of the biggest AI companies ⁓ in the world. And for me, a lesson here is clear is that supply chain security is absolutely critical. ⁓ and especially in the AI space where open source dependencies are absolutely everywhere. And I think secondly, if you're building AI systems, look at treating your data pipeline with the same rigor as you would your production infrastructure. ⁓ And this is also a brilliant experience of why having governance matters. Because for me, you can't just adopt open source tools without understanding the risks that they introduce. And this is exactly the kind of AI security convergence issues that we focus on at the AI and cybersecurity association because they're just so deeply intertwined.
Anmol: And we had Yesenia Yser on the show a couple of weeks ago, she specializes in open source AI security. She's contributed to the Linux Foundation. And one thing we talked about there, and Lisa, you might have a comment, the idea that sometimes in open source, there are so many dependencies and so many different things happening in those open source repositories that difficult to figure out where exactly that third party risk comes from. do you have any additional advice for people who do integrate open source because they don't have access to the data that they need, right? Because oftentimes getting data is a challenge. ⁓ then people use open source data sets. But do you have any advice as far as ⁓ they could determine open source risk?
Lisa Ventura MBE FCIIS: Absolutely, this is a great point ⁓ and it's something a lot of our organizations really struggle with and I've seen that firsthand. ⁓ Mainly because with traditional third party risk, you have a vendor, you have a contract, you can send a security questionnaire or ask for their SOC 2 report, but with open source dependencies, it's completely different because you might have a developer. pulling in a library that's got lots of its own dependencies underneath it. And nobody within the organization often has consciously evaluated any of them. And this is why the Merc or Breach that I talked about is a great example of this because Light LRM is a popular, widely used project and many companies relied on it. But I do wonder how many of them had actually assessed the security posture. of that project in terms of who maintains it, how contributions are reviewed, or what would happen if it were actually compromised. So practically, here's a few things I think that organizations can do. I'd say maintain a software bill of materials, and that's sometimes called an SBOM. And that gives you visibility into what open source components you're actually using across your estate. because I think you can't manage risk that you can't see. And then also use tools that monitor for known vulnerabilities ⁓ in your dependencies. So think about things like Dependabot or Cynic or similar platforms that can flag those issues automatically when the vulnerability is disclosed in something that you rely on. And I think for me, this is where it gets a bit more cultural. So build open source risk into your existing third party risk framework. And just because there's no contract doesn't mean that there's no risk. For me, the absence of a contractual relationship arguably makes it actually higher risk because there's no obligation on anyone to fix things when they go wrong or even to notify you if something goes wrong as well. And finally, I would say have a response plan. So if critical open source dependency is compromised tomorrow, does your team know what to do and who in your team would own that decision? How quickly can you identify what systems are effective? And that often is the preparedness piece that's often missing. ⁓ And I'll just finish this off by saying, I'm not saying avoid anything open source or don't use it because it's brilliant and it powers the vast majority of modern software, but I think just be a bit more intentional informed about what you're bringing into your environment when you're looking to deploy anything that's open source related.
Anmol: Yeah, great points there. So Lisa, one interesting point you brought up that I haven't heard before is the idea of the contractual agreement, because that I think is something people have not really talked about. So typically with open source is just you adopt it. And then if you're a business, you're the one who has to be paying attention to the open source repositories. Is it being maintained properly? Was there a CVE that was detected? Right? Was there a critical vulnerability that you need to patch right away. But that's something typically the businesses when they use these open source tools, a lot of times they might just assume, okay, it works because all of these other companies are using it or is very popular, or it has a good reputation in the industry. But it's important to realize that because it's open source, you don't have a contractual agreement if something goes wrong. someone might not fix that issue because it's open source, it's volunteers working on this. So you need to also have your own team in place to be aware of when something goes wrong. What is your incident response plan?
Lisa Ventura MBE FCIIS: Absolutely, and that's what I mentioned as well, that incident response plan is critical and not just having that plan, but testing that plan, ⁓ perhaps running it out with some tabletop exercises, some real world scenarios, things that could potentially go wrong and test those incident response plans as well for sure.
Anmol: Yeah, and I would say test it in a way that is realistic. So sometimes people have hundreds of pages of incident response plans, but in the event of an emergency, you won't follow a hundred page document. It will be like a quick, small set of steps. It won't be hundreds of pages, especially accounting for what happens when people are stressed and there are stressful situations. People would behave differently than a calm, drill, you know, like a fire alarm drill. When it's a drill, everyone knows how to walk in line and evacuate the building. But when there is an actual fire, people might panic. And the same thing with incident response in cybersecurity.
Lisa Ventura MBE FCIIS: Absolutely, and I'd love to just come in and something you mentioned with regard to the stressful, you know, situations and the high precedence of burnout that we see in the sector. And for me, it's absolutely no coincidence that we have a huge problem with mental health stress and burnout. Yeah, we're online a lot of the time 24 seven, we're constantly monitoring for the next incident, the next threat, the next vulnerability that needs to be plugged or patched. ⁓ And for me as well, particularly here in the in the UK, last year, we seemed to be absolutely hammered with cyber attacks. ⁓ It's almost a ⁓ year on since Marks and Spencer's, one of our big retail flagship brands, was hit by a huge scale cyber attack, ⁓ followed by a myriad of others. And we had the Co-op Group, had ⁓ Jaguar Land Rover, we had some other luxury brands like Pandora, Cartier, etc. They were all hit. Harrods was hit. ⁓ And there just seemed to be one after the other after the other. And for me, while the media was all reporting on the breach and all the downtime and everything that had happened, I'm thinking, I hope everyone that's involved in fixing these things and getting everything back up and running are being looked after. Because that was the thing that was missing for me, that there was no real emphasis on all those people that have been working round the clock to get everything back up and running again. So it's something that I feel really strongly about ⁓ because in our field, the threat just never sleeps. It's there 24-7 all year round. And there is that constant pressure to stay ahead. And we've seen in the media the consequences of getting it wrong that can be severe. And just touching on the Jaguar Land Rover cyber attack that we had here in the UK at the end of August last year. They couldn't just restart their production. It was a very phased approach and it took weeks. I actually went on the national news over here and talked about it and I said at the time that it happened that it's going to take weeks if not months for production to get restarted again and that's exactly how it played out. In the meantime, some smaller businesses that relied on Jaguar Land Rover struggled massively and one or two actually went out of business. Classic example there, there was a company that have a fleet of coaches that ferries all the workers to the different car production sites and back and forth into Birmingham. And that company had invested in a fleet of new coaches. spent a lot of money on getting all this infrastructure in place. The cyber attack happened. Suddenly they're not ferrying people back and forth to the production lines. And within no time at all, they went out of business. So that's the first hand consequences of an attack of that nature. And what I will say is look out for your employees. first and foremost, check in on them if you are going through a live situation. ⁓ Make sure that they're looked after, you check in on them ⁓ and make sure that burnout is something that you address within your organisation.
Anmol: instead of ⁓ focusing on the hype the media, just focusing on the basic cybersecurity principles is so important. Have subject experts who know the AI part data scientists, ⁓ because sometimes you need to know how the models so that when they're attacked, you know how to recognize an attack. Because even if you use generative AI to detect these attacks, it can make mistakes. There's no model that's 100 % accurate. It can hallucinate or ⁓ think of things that aren't there. ⁓ But it's important to focus on the ⁓ cybersecurity principles, basics, having your basics covered, like Lisa mentioned, All of those things, incident response, evaluating any third party risks, supply chain security, and then investing in your people, which Lisa, I think that's a really important point because... all of this AI security, all comes down to the people. We have people on all sides, people that are defending, people that are attacking. And a lot of times, I would say our field is not just the technical side, but it's also the human side and the psychology side as well.
Lisa Ventura MBE FCIIS: Absolutely, and that's my area that I work in day in and day out. And I see so much emphasis, rightly so, don't get me wrong, on the technical side. But the human factors are equally, not just as important as the technical side. So do make sure you check in with your people and make sure that they're not at risk of burnout and that you look after them.
Anmol: Thanks Lisa for that point. So my next question was, if you think about ⁓ we've talked about, is there an everyday scenario that comes to mind that explains one of these concepts to the audience?
Lisa Ventura MBE FCIIS: Absolutely. So I'm going to use one that I hope that everyone can relate to. your morning coffee run. So you might walk into your favourite coffee shop, you connect to the free Wi-Fi, you might open your phone, you might check your emails, ⁓ maybe even quickly log into a work app and that feels completely routine. Now imagine that there's an attacker sitting in that same coffee shop and with AI powered tools, they can set up a fake Wi-Fi hotspot that looks identical to the coffee shop's real one. They can then use AI to create a convincing captive portal that mirrors the coffee shop's branding perfectly. And once you connect, they can intercept your traffic, capture your login credentials. or even serve you a phishing page that looks exactly like your email login. But here's the bit that's changed because AI now let's let them do all of this at scale automatically and with that level of polish as it were that would have taken a skilled attacker hours and hours to set up manually. So I think the barrier to entry has dropped dramatically with this. So. In terms of this particular scenario, what you can do, always, always use a VPN when you're logged on to public Wi-Fi and be cautious about which networks you're connected to. Don't log into any ⁓ sensitive or corporate or work accounts on any open networks. And if your organisation hasn't got this already, push for a policy that covers how employees should handle connectivity. when they're out and about. So perhaps a bit like a remote access policy. ⁓ All of these aren't complicated things, but they can make a really big difference.
Anmol: Yeah, thank you so much, Lisa, for that explanation. think that was a great example, because most of us, drink coffee or tea in the morning, you so we can stay awake for work. So that's definitely a relatable example there. thank you so much Lisa for being on this show. Are there any resources you recommend for our audience to learn more?
Lisa Ventura MBE FCIIS: Absolutely. So ⁓ please visit the AI and Cybersecurity Association website and see the work that we're doing as we gear towards launch, ⁓ because there will be opportunities for people in the US and internationally to join ⁓ as international members as well. ⁓ My current book, Artificial Intelligence and Cybersecurity by Kogan Page, ⁓ that covers the landscape between AI and cybersecurity in depth. from how it's being used by attackers and defenders ⁓ to governance and ethics. ⁓ And one of the couple of recent reports actually that I'll highlight ⁓ CrowdStrike's global threat report I thought was excellent for this year as was Kiteworks state of AI cybersecurity report ⁓ because they had tons of real data ⁓ in them rather than just ⁓ sort of industry, FUD and hype, know, fear, uncertainty and doubt is, I'm sure we've known. And for anyone that's listening as well, I also run a monthly community event on Zoom here with the UK called InfoSet Lunch Hour. And that takes place on the first Wednesday of every month at 12.30 PM UK time. But we've covered things like AI biases, supply chain security. neurodiversity and cyber, know, all sorts and everyone's welcome. It's run under Chatham House rules and it's a great way to meet others network in the industry and get some great insights over lunch at the same time.
Anmol: Yeah, that InfoSec lunch hour is really great. But for those of us in the US, it's more like breakfast hour. it's like basically 7.30 Eastern time. So for US, think of it like a breakfast, InfoSec breakfast hour. But yeah, it's a really, really great group. And Lisa, you've done so much for the industry and you're so inspiring. So thank you so much for being on the show. Is there anything else you want to add before we end the episode?
Lisa Ventura MBE FCIIS: Yeah, so it's too early. Yep. Just to say thank you for having me and if anybody listening, you'd like to get in touch with me, my website is lisaventura.co.uk or you can find me on LinkedIn. ⁓ pop me a request or a connection request or send me a message. ⁓ love to talk to you all.
Anmol: thank you so much, Lisa. This concludes our episode of the AI Security Update. We'll be back next time with more great conversations. ⁓
Lisa Ventura MBE FCIIS: It's absolutely wonderful to be here today. Thank you for having me.
Anmol: Lisa, can you introduce yourself a little bit more on your background and your experience related to AI and cybersecurity?
Lisa Ventura MBE FCIIS: Absolutely. So as you rightly mentioned, I've published books, I founded Cybersecurity Unity some years ago, but more recently, I've actually founded what I believe to be the world's first membership body and trade association that's solely dedicated to the intersection of AI and cybersecurity. and that's called the AI and cybersecurity association, through which we'll lobby government for the safe, secure, responsible and ethical use of AI, raise awareness of the use of AI. ⁓ I guess not so much for the huge conglomerates, the big businesses, ⁓ but more for the smaller businesses that may be using it and especially using a lot of the learning language models day to day. but not sort of really thinking about the security ramifications around those. So that's very much ⁓ where my work is going to be focused on going forward. And I've also had another book published with Kogan Page, which is called Artificial Intelligence in Cybersecurity. And that covers how AI is transforming the threat landscape and how we defend against it. And it was during that process of writing the book and researching the book that I had the idea for the trade association. ⁓ Everything I do as well around the cybersecurity awareness training space, it's always empowerment led, trauma informed, and importantly for me, neuro inclusive, because I'm also openly neurodivergent. And I was diagnosed with autism and ADHD a few years back.
Anmol: Thank you so much, Lisa. We also had Nathan Chung on our show, who's also openly neurodivergent and he's a neurodiversity champion. And I think that brings a unique angle to how approach problems and solve problems. So ⁓ for cybersecurity, we need all skill sets. ⁓ there anything else you want to add about neurodiversity or ⁓
Lisa Ventura MBE FCIIS: I mean, me, I've done a lot of work ⁓ research that area and also act as a mentor and support people that are neurodivergent to get into careers in ⁓ cybersecurity. many studies have shown that those of us who are neurodivergent are actually really well suited to careers ⁓ within cybersecurity space, ⁓ mainly because a lot of the time we have strong pattern recognition, have strong hyper focus and so on. And that lends itself really well to many different roles within the cybersecurity space.
Anmol: Yeah, and I would say that also applies to AI security as well. When you look at the different attacks, one of the classic detections is anomaly detection, right? So seeing what doesn't match the pattern. And that, again, is pattern recognition, which is very important.
Lisa Ventura MBE FCIIS: sure. Mm-hmm.
Anmol: based on your experience, what challenges have you seen related to AI security? ⁓ You mentioned small businesses, but what are the challenges particular?
Lisa Ventura MBE FCIIS: I've ⁓ quite a few and that led me to the founding of the AI and Cybersecurity Association with my research for my recent book. ⁓ I think for me, it's the huge rush ⁓ I see organizations have to adopt this technology. And a lot of the time I'll sit down with them and they'll say to me, we need to roll out copilot or Gemini or whatever tool they've chosen. ⁓ And I'll say to them, but what specific business case do you have for the adoption of this technology? And what are you looking for AI to solve in your business? And they'll turn around and say, well, we don't know, but everybody's got it. So we need to adopt it. We need to have it. And there is this huge rush to adopt this technology. ⁓ And for me, I do a lot of awareness raising to businesses to say, It's great that you want to adopt this, but make sure you have those guardrails in place. Make sure you have those AI policies in place. Make sure as well you adopt and use the tools that will be the most benefit to your business and your organization. And really importantly as well, have mechanisms in place ⁓ to have approved tools for use within your organization. and also be aware of any shadow AI, is becoming a bigger problem in shadow IT, I think at the moment. And be aware of that and that people might be using different tools almost sort of underground. I mean, I've seen organizations that have just gone a blanket ban, no, we're not using any of it. And that's accelerated their shadow AI. So... For me, those are certainly some of the challenges that I'm seeing in the space. Also from an AI and security perspective, a lot of people that are involved in the cyber awareness and training arena, it's rendering a lot of the traditional advice that we would often give around spotting what might potentially be phishing email simulations and so on. AI deep fake videos, they're getting so, so much better and so much more convincing. And a lot of that traditional advice of look at the spelling, look at the grammar, look at how an email is constructed and then so on. With attackers using these models to produce these emails, you the average person on the street will look at it and potentially think that it's a genuine ⁓ email when in fact it's a phishing attempt. So those are certainly some of the biggest challenges that I'm seeing in the space at the moment.
Anmol: you need to think the use case and what your business actually does and analyze and see what data do you have? do you actually need AI to process the data? I always say sometimes a manual approach is better than AI if you can use a traditional spreadsheet. Let's say you have like 10 data points. You don't need AI for that. AI is very good for like thousands and thousands of data points, a lot of data points, you're trying to find one needle in a haystack ⁓ that's not something a person can do, then maybe try to use AI. But it requires a lot of analysis and determining whether or not you really need to use AI for your use case or for your business. So definitely agreed there. And Lisa, we see AI and AI security in the news. People hear about it almost every day, and there's always a new development. based on what you've seen in the news, whether it's global news or local news to you, what a current event that you think our audience should know about?
Lisa Ventura MBE FCIIS: Absolutely. So for me, there was quite a significant one that broke ⁓ a couple of weeks ago, and that was actually the Merkle ⁓ So ⁓ is actually an AI recruiting startup, and it was hit with a supply chain cyber attack, ⁓ and it involved ⁓ open source project Lite LLM. ⁓ for anyone that's not ⁓ with this, Merkle recruits domain experts such as scientists, doctors and lawyers. ⁓ And they also then provide training data for major AI companies and their customers ⁓ actually include Anthropic, OpenAI ⁓ and Meta as well. ⁓ for me with this, we're actually talking about a company that sits right at the heart of the AI ecosystem. And what happened with this particular breach is that a hacking group called Team PCP, it actually compromised Light Allalem. And that's actually a widely used open source library for connecting applications to AI services. ⁓ Merkle was actually one of lots of companies that were affected by this. ⁓ then separately, another extortion group claimed responsibility and shared samples ⁓ stolen data, which reportedly included information from Slack, support ticket data, and even recordings of conversations between Merkle's AI systems and contractors. And for me, this really matters a lot because according to unconfirmed reports, data sets used by some of Merkle's customers and information, especially around some quite secretive ⁓ AI projects. may have been compromised. So the implications for this are huge because you compromise one link in the supply chain and that could potentially then grant access to the training data and project details of some of the biggest AI companies ⁓ in the world. And for me, a lesson here is clear is that supply chain security is absolutely critical. ⁓ and especially in the AI space where open source dependencies are absolutely everywhere. And I think secondly, if you're building AI systems, look at treating your data pipeline with the same rigor as you would your production infrastructure. ⁓ And this is also a brilliant experience of why having governance matters. Because for me, you can't just adopt open source tools without understanding the risks that they introduce. And this is exactly the kind of AI security convergence issues that we focus on at the AI and cybersecurity association because they're just so deeply intertwined.
Anmol: And we had Yesenia Yser on the show a couple of weeks ago, she specializes in open source AI security. She's contributed to the Linux Foundation. And one thing we talked about there, and Lisa, you might have a comment, the idea that sometimes in open source, there are so many dependencies and so many different things happening in those open source repositories that difficult to figure out where exactly that third party risk comes from. do you have any additional advice for people who do integrate open source because they don't have access to the data that they need, right? Because oftentimes getting data is a challenge. ⁓ then people use open source data sets. But do you have any advice as far as ⁓ they could determine open source risk?
Lisa Ventura MBE FCIIS: Absolutely, this is a great point ⁓ and it's something a lot of our organizations really struggle with and I've seen that firsthand. ⁓ Mainly because with traditional third party risk, you have a vendor, you have a contract, you can send a security questionnaire or ask for their SOC 2 report, but with open source dependencies, it's completely different because you might have a developer. pulling in a library that's got lots of its own dependencies underneath it. And nobody within the organization often has consciously evaluated any of them. And this is why the Merc or Breach that I talked about is a great example of this because Light LRM is a popular, widely used project and many companies relied on it. But I do wonder how many of them had actually assessed the security posture. of that project in terms of who maintains it, how contributions are reviewed, or what would happen if it were actually compromised. So practically, here's a few things I think that organizations can do. I'd say maintain a software bill of materials, and that's sometimes called an SBOM. And that gives you visibility into what open source components you're actually using across your estate. because I think you can't manage risk that you can't see. And then also use tools that monitor for known vulnerabilities ⁓ in your dependencies. So think about things like Dependabot or Cynic or similar platforms that can flag those issues automatically when the vulnerability is disclosed in something that you rely on. And I think for me, this is where it gets a bit more cultural. So build open source risk into your existing third party risk framework. And just because there's no contract doesn't mean that there's no risk. For me, the absence of a contractual relationship arguably makes it actually higher risk because there's no obligation on anyone to fix things when they go wrong or even to notify you if something goes wrong as well. And finally, I would say have a response plan. So if critical open source dependency is compromised tomorrow, does your team know what to do and who in your team would own that decision? How quickly can you identify what systems are effective? And that often is the preparedness piece that's often missing. ⁓ And I'll just finish this off by saying, I'm not saying avoid anything open source or don't use it because it's brilliant and it powers the vast majority of modern software, but I think just be a bit more intentional informed about what you're bringing into your environment when you're looking to deploy anything that's open source related.
Anmol: Yeah, great points there. So Lisa, one interesting point you brought up that I haven't heard before is the idea of the contractual agreement, because that I think is something people have not really talked about. So typically with open source is just you adopt it. And then if you're a business, you're the one who has to be paying attention to the open source repositories. Is it being maintained properly? Was there a CVE that was detected? Right? Was there a critical vulnerability that you need to patch right away. But that's something typically the businesses when they use these open source tools, a lot of times they might just assume, okay, it works because all of these other companies are using it or is very popular, or it has a good reputation in the industry. But it's important to realize that because it's open source, you don't have a contractual agreement if something goes wrong. someone might not fix that issue because it's open source, it's volunteers working on this. So you need to also have your own team in place to be aware of when something goes wrong. What is your incident response plan?
Lisa Ventura MBE FCIIS: Absolutely, and that's what I mentioned as well, that incident response plan is critical and not just having that plan, but testing that plan, ⁓ perhaps running it out with some tabletop exercises, some real world scenarios, things that could potentially go wrong and test those incident response plans as well for sure.
Anmol: Yeah, and I would say test it in a way that is realistic. So sometimes people have hundreds of pages of incident response plans, but in the event of an emergency, you won't follow a hundred page document. It will be like a quick, small set of steps. It won't be hundreds of pages, especially accounting for what happens when people are stressed and there are stressful situations. People would behave differently than a calm, drill, you know, like a fire alarm drill. When it's a drill, everyone knows how to walk in line and evacuate the building. But when there is an actual fire, people might panic. And the same thing with incident response in cybersecurity.
Lisa Ventura MBE FCIIS: Absolutely, and I'd love to just come in and something you mentioned with regard to the stressful, you know, situations and the high precedence of burnout that we see in the sector. And for me, it's absolutely no coincidence that we have a huge problem with mental health stress and burnout. Yeah, we're online a lot of the time 24 seven, we're constantly monitoring for the next incident, the next threat, the next vulnerability that needs to be plugged or patched. ⁓ And for me as well, particularly here in the in the UK, last year, we seemed to be absolutely hammered with cyber attacks. ⁓ It's almost a ⁓ year on since Marks and Spencer's, one of our big retail flagship brands, was hit by a huge scale cyber attack, ⁓ followed by a myriad of others. And we had the Co-op Group, had ⁓ Jaguar Land Rover, we had some other luxury brands like Pandora, Cartier, etc. They were all hit. Harrods was hit. ⁓ And there just seemed to be one after the other after the other. And for me, while the media was all reporting on the breach and all the downtime and everything that had happened, I'm thinking, I hope everyone that's involved in fixing these things and getting everything back up and running are being looked after. Because that was the thing that was missing for me, that there was no real emphasis on all those people that have been working round the clock to get everything back up and running again. So it's something that I feel really strongly about ⁓ because in our field, the threat just never sleeps. It's there 24-7 all year round. And there is that constant pressure to stay ahead. And we've seen in the media the consequences of getting it wrong that can be severe. And just touching on the Jaguar Land Rover cyber attack that we had here in the UK at the end of August last year. They couldn't just restart their production. It was a very phased approach and it took weeks. I actually went on the national news over here and talked about it and I said at the time that it happened that it's going to take weeks if not months for production to get restarted again and that's exactly how it played out. In the meantime, some smaller businesses that relied on Jaguar Land Rover struggled massively and one or two actually went out of business. Classic example there, there was a company that have a fleet of coaches that ferries all the workers to the different car production sites and back and forth into Birmingham. And that company had invested in a fleet of new coaches. spent a lot of money on getting all this infrastructure in place. The cyber attack happened. Suddenly they're not ferrying people back and forth to the production lines. And within no time at all, they went out of business. So that's the first hand consequences of an attack of that nature. And what I will say is look out for your employees. first and foremost, check in on them if you are going through a live situation. ⁓ Make sure that they're looked after, you check in on them ⁓ and make sure that burnout is something that you address within your organisation.
Anmol: instead of ⁓ focusing on the hype the media, just focusing on the basic cybersecurity principles is so important. Have subject experts who know the AI part data scientists, ⁓ because sometimes you need to know how the models so that when they're attacked, you know how to recognize an attack. Because even if you use generative AI to detect these attacks, it can make mistakes. There's no model that's 100 % accurate. It can hallucinate or ⁓ think of things that aren't there. ⁓ But it's important to focus on the ⁓ cybersecurity principles, basics, having your basics covered, like Lisa mentioned, All of those things, incident response, evaluating any third party risks, supply chain security, and then investing in your people, which Lisa, I think that's a really important point because... all of this AI security, all comes down to the people. We have people on all sides, people that are defending, people that are attacking. And a lot of times, I would say our field is not just the technical side, but it's also the human side and the psychology side as well.
Lisa Ventura MBE FCIIS: Absolutely, and that's my area that I work in day in and day out. And I see so much emphasis, rightly so, don't get me wrong, on the technical side. But the human factors are equally, not just as important as the technical side. So do make sure you check in with your people and make sure that they're not at risk of burnout and that you look after them.
Anmol: Thanks Lisa for that point. So my next question was, if you think about ⁓ we've talked about, is there an everyday scenario that comes to mind that explains one of these concepts to the audience?
Lisa Ventura MBE FCIIS: Absolutely. So I'm going to use one that I hope that everyone can relate to. your morning coffee run. So you might walk into your favourite coffee shop, you connect to the free Wi-Fi, you might open your phone, you might check your emails, ⁓ maybe even quickly log into a work app and that feels completely routine. Now imagine that there's an attacker sitting in that same coffee shop and with AI powered tools, they can set up a fake Wi-Fi hotspot that looks identical to the coffee shop's real one. They can then use AI to create a convincing captive portal that mirrors the coffee shop's branding perfectly. And once you connect, they can intercept your traffic, capture your login credentials. or even serve you a phishing page that looks exactly like your email login. But here's the bit that's changed because AI now let's let them do all of this at scale automatically and with that level of polish as it were that would have taken a skilled attacker hours and hours to set up manually. So I think the barrier to entry has dropped dramatically with this. So. In terms of this particular scenario, what you can do, always, always use a VPN when you're logged on to public Wi-Fi and be cautious about which networks you're connected to. Don't log into any ⁓ sensitive or corporate or work accounts on any open networks. And if your organisation hasn't got this already, push for a policy that covers how employees should handle connectivity. when they're out and about. So perhaps a bit like a remote access policy. ⁓ All of these aren't complicated things, but they can make a really big difference.
Anmol: Yeah, thank you so much, Lisa, for that explanation. think that was a great example, because most of us, drink coffee or tea in the morning, you so we can stay awake for work. So that's definitely a relatable example there. thank you so much Lisa for being on this show. Are there any resources you recommend for our audience to learn more?
Lisa Ventura MBE FCIIS: Absolutely. So ⁓ please visit the AI and Cybersecurity Association website and see the work that we're doing as we gear towards launch, ⁓ because there will be opportunities for people in the US and internationally to join ⁓ as international members as well. ⁓ My current book, Artificial Intelligence and Cybersecurity by Kogan Page, ⁓ that covers the landscape between AI and cybersecurity in depth. from how it's being used by attackers and defenders ⁓ to governance and ethics. ⁓ And one of the couple of recent reports actually that I'll highlight ⁓ CrowdStrike's global threat report I thought was excellent for this year as was Kiteworks state of AI cybersecurity report ⁓ because they had tons of real data ⁓ in them rather than just ⁓ sort of industry, FUD and hype, know, fear, uncertainty and doubt is, I'm sure we've known. And for anyone that's listening as well, I also run a monthly community event on Zoom here with the UK called InfoSet Lunch Hour. And that takes place on the first Wednesday of every month at 12.30 PM UK time. But we've covered things like AI biases, supply chain security. neurodiversity and cyber, know, all sorts and everyone's welcome. It's run under Chatham House rules and it's a great way to meet others network in the industry and get some great insights over lunch at the same time.
Anmol: Yeah, that InfoSec lunch hour is really great. But for those of us in the US, it's more like breakfast hour. it's like basically 7.30 Eastern time. So for US, think of it like a breakfast, InfoSec breakfast hour. But yeah, it's a really, really great group. And Lisa, you've done so much for the industry and you're so inspiring. So thank you so much for being on the show. Is there anything else you want to add before we end the episode?
Lisa Ventura MBE FCIIS: Yeah, so it's too early. Yep. Just to say thank you for having me and if anybody listening, you'd like to get in touch with me, my website is lisaventura.co.uk or you can find me on LinkedIn. ⁓ pop me a request or a connection request or send me a message. ⁓ love to talk to you all.
Anmol: thank you so much, Lisa. This concludes our episode of the AI Security Update. We'll be back next time with more great conversations. ⁓
