Podcast: Cybersecurity concerns maintenance professionals often overlook
Key Highlights
- OT cybersecurity is a reliability issue: Infected devices, compromised PLCs and faulty code can trigger production outages and equipment failures.
- Use AI to strengthen proactive defense: AI can identify vulnerabilities, uncover overlooked maintenance risks and help prioritize remediation before incidents occur.
- Build contextual asset visibility: Document every digital asset, including legacy and hidden connections, and assess criticality, redundancy and operational impact to prioritize security investments.
- Make cybersecurity a business priority: Quantify risks in terms of production downtime, revenue, safety and regulatory exposure to build the case for upgrades, redundancy and workforce knowledge retention.
On this episode of Great Question we spoke to two guests: Honeywell Process Automation Portfolio Director Paul Smith and Rockwell Automation Global Director of Cybersecurity Services Rick Kaun. Our discussion covered how to integrate IT and OT cybersecurity practices and the current state of cybersecurity as applied to maintenance and reliability.
Below is a transcript of the episode:
PS: My name is Paul Smith. I'm the portfolio director for Honeywell Process Automation, basically taking care of all the commercial product offerings that we have, so products and services. Some of that stuff that encompasses is secure operations center. We have cyber insights, proactive defense, GRC, and then assessment and field services, and that can span anything from reliability to incident response. IEC 62443 assessments, pen testing, tabletops, purple teaming, all that, all that fun stuff, bubbles up to me for my ownership.
MC: So I want to start in a very broad sense with why is OT cybersecurity important to maintenance and reliability?
PS: Oh, start with hard questions. So why is it important? The importance behind cybersecurity in the maintenance and reliability sector is essentially this: anything that can be seen as a threat or an impact to a process should be on the purview of a reliability engineer.
And if I can walk in with an infected laptop, that's actually going to cause an outage by locking out a controller or bricking a a PLC, these are things that can happen from a normal day activity for someone pushing a wrong configuration into a controller, PLC, RTU, flow computer. If you were using a flow computer, gas chromatograph. That's a smart gas chromatograph. Whatever happens to be just out of sheer, sheer issues when it comes to developing the wrong logic, you can do that from a third-party engineering firm. You can do it from yourself. Sometimes you fat finger something in the ladder logic, and then it comes, and it doesn't get fully tested through your lab. You can put it out into the field now. That gets magnified or amplified when you start to wrap cyber around it.
Cyber is super critical because when people can start injecting things into your codebase, it's no longer you creating the problem by having like a mental moment and and putting in some wrong rungs. Someone can actually inject rungs into the program before it goes out. So all these things have to be super vetted and analyzed and aware that anything is possible now with the velocity of things happening in the AI space, the barrier of entry in the past, like the barrier entry has been lowered. Like in the past, you would have to have people who are specialists in this field. You'd have to have someone who's like an automation specialist, a process control expert that also understands the IT cyber part of the business, and then be able to coalesce that altogether.
Now you can go to whatever prompt you want, and ask it basically the stepping stones that you need to be malicious. At one point, 4 years ago, when everything was in its infant stage, you could just get, you could ask for an exploit to be auto-generated for you. While now a lot of the big ones like OpenAI and Anthropic, Nemotron from Nvidia are starting to put guardrails into the prompt, so prompt injection won't actually produce maliciousness. But there's still the barrier of entry has been lowered. I digress. That was a long ramble on why I feel cyber is important because it just has such a great impact.
MC: I think it's of course very timely that you include AI as part of that conversation. And so the report, and that is the OT Security Benchmark Report from Honeywell, says that 72% of OT security professionals use AI-enabled threat detection, and that OT security professionals must determine how much authority to give to AI software. And as you pointed out, we have AI on the the threat actor side, and we have AI on the defender side, and we have. AI in just people are asking questions on their phone while they're working. So, what are the pros and cons of using AI in reliability and maintenance?
PS: When AI first came out, people asked 'what is this going to do for us?' And I said it was the next generation of search engine. So, in its infancy, when you go to look at Google and you ask for some results from Google, it'll give you like the top 10, and then some of them it's messy. And depending on who pays what for promotion, you'll get their websites. You had to do your own research and basically dig into the websites that find the correct context. With AI, when it first came out, you could actually say ‘I want to know the results’ and it would just generate the outcome that you were expecting. So that's where I really parallel the analogy to being the next search engine.
Fast forward – now, I want to clarify AI is the fastest. Like we have not experienced this in our lifetimes, the evolution of how fast AI is growing. They're saying every six months it's exponentially expanding and doubling in in complexity and outcomes. We are creating a super intelligence that we don't – sorry, I don't want to take this into an AI talking point. But basically, I mean, if you're not leveraging it to empower your day-to-day activities, you're really losing out.
Because I don't feel that AI is going to displace people. I think people who use AI will displace people. That makes sense. It’s a tool in your tool bag that you need to be using. You can leverage it for that next-generation search engine, but also there's certain stuff inside your environment. Once you start letting it look at your data, it's going to find holes in some of your logic and theory that you've been training on.
Like even if you're a 30 year veteran and you're looking at your ladder logic. You're looking at your maintenance schedules. You're looking at all this data. There are parts of that context and telemetry from data points that you might miss that leveraging AI will be able to stitch together and give you a better result. So if there's like a piece that for whatever reason got missed in your your build out, then it will automatically pull it in, and then you can reconstruct how your your go-forward plan will happen. Super advantageous. See it time and time again. There's always that one nugget that it yanks in from all of its contextual awareness and knowledge from its training data that I always leverage and start to see, like, oh yeah, I forgot to add that that piece into my decision path.
MC: So, I'm curious with this next question whether the answer is actually the same as what you just mentioned, but almost approached from the opposite direction. The report mentions that legacy systems, whether those are legacy operating systems or disconnected machines, can be a major point of vulnerability. How can this complicate cybersecurity for maintenance and reliability? And what tools are available to respond to this? I think you might have answered that, but anything you want to add about those legacy systems, things that haven't been updated, things that are not connected at all, or you wouldn't think they're connected?
PS: I give grace and empathy to business owners that have to manage these systems. Engineers that have to manage them that ask for budget for replacing, but they have to wait. As a business owner, you have to wait. Well, what's the risk versus the investment? And if the risk is low, the investment's high. For instance, there is a company, and I'm trying to make sure I don't divulge too much.
There's a company that has an like a legacy infrastructure for a parking system, basically those parking systems that tell you if there's a free space in their their parkade, and it's run on an old I think Windows. I think they upgraded it to a Windows 7 box to run the software, but just the infrastructure itself to rip and replace was in the millions of dollars, and they were saying: this system is high risk because we can't patch it. It's legacy, but we feel it's disconnected. However, we do manage it centrally from another location. So, by just the definition, when I would talk to them, I'm like, it's not air gapped because you do have a there is connectivity between this Windows 7 device and a central system that is connected to your broader organization. Now, with that said, that point of entry, that Windows 7 box, is in a closet in the parkade where anyone can open that door. Because it's not a locked door. Anyone can open that door or see this dusty old closet and and stick in a USB key that can backhaul it into their organization.
So it's stuff like that where you don't think ... it’s a computer that's controlling a parking lot lights in a system to show you which which spots are available? Which spots are taken? But it has this sneaky backdoor into the broader organization, and that is stuff that has to be documented and well understood. And see, like on a risk assessment, when you're doing the analysis, what's the blowback from this? Like, what do we have to put in place to segment this? Can we segment off? Can we put a diode possibly in between the two? When you're saying it's connected to a central management system, is it more for visibility and less and less from a control? If there's no control that needs to be done, so I don't have to put any updates to that system. I don't have to ... we're never going to change it because just the cost of changing and the cost of ownership is way too high. Then can we make sure we're only reading the data specific to the outcomes that we expect, and we're not opening ourselves up for a a broader breach because of simple asset management. So it's really understanding like what that exposure is, what's the outcome expected from a system, and if we can put in any device around it. Now, finding those rogue systems is a harder task, and going through like if you don't know what you've inherited.
So the customers that inherited this parkade when they bought the complex. They didn't know they had this. We went through an audit, and we just happened to find it in an audit, matter of factly. And we found out that okay, well, what's this cable? Why is it connected to a switch? And what does that switch go to? And we had to go and trace it. I mean, that took a week or so. Finally, finding out where it was and that it had a full TCP handshake between the two. There was no firewall in between it, and then that turned into a larger remediation.
But I mean, it came down to tools not being able to fully fingerprint because even in that situation we put tools out there and we saw there was a Windows 7 box, but that Windows 7 box was just hanging off a switch and we couldn't find it, so we had to go walk and find where this this Windows host was, right? So, and it took weeks to to walk through everything and try to find what the system was. So, I empathize with the business owners that have to replace these legacy systems, and then I also empathize with the engineers that have to support and maintain them because it is a conundrum.
MC: I respect how much work goes into that, and also how many of those conversations do come down to budget. And as you said, it's risk versus the benefit of that. So in maintenance and reliability, a lot of times the primary software they're working with is a CMMS, that computerized maintenance management system. Any cybersecurity considerations specifically for that CMMS?
PS: So, on the CMMS, it depends. It depends on dependency. So once again, if we start to pull through on the software development side, right? Because that's where you always find the vulnerabilities hidden in the code, is where someone has actually implemented. And I'm going to stay away from more of the connectivity of the backend system and talk about the software itself. Because where you really find it is on dependencies that have been just pulled in, and we saw this in a lot of the development here lately with NPM and poison packages out there on the software development. You have legacy systems that have intrinsically vulnerable components that haven't been patched or won't be patched. The cost to patching them.
This is honestly on the maintenance systems are a little bit more of an uplift to do it. You are inheriting usually code that was developed and has … I'm trying to be sensitive because I know all the the manufacturers, and I don't want to call anyone out … You're in some legacy code base, and to get something new is usually a cost a cost benefit that they have to figure out if there's ROI in doing a migration to a new system. So you're always sitting there with these legacy dependencies that are intrinsically in the code that someone can exploit.
And the reason why I brought up NPM because it's one of the biggest ones that came out. Where it's so easy for a developer in a period of time – it gets worse when we pull AI into software development – but in a period of time, you would have developers just doing NPM install, or, I'm going to install these packages, and inside those packages are crypto keys, security keys. There you have dependencies that are just riddled with vulnerabilities that are all sitting embedded in some sort of dependency function or program that's rolled up to the system, that as soon as someone launches a piece of code to exploit it, they have full control of the system.
Now, depending on some of the protocols that communicate back, and depending on the systems that you have, you can actually you can fire off wrapped protocol packets that will actually cause damage to those systems as well, because it will tell everything to shut down via one protocol payload. I will give one example. There is a product out there that you … and I'll say it here: function code 90. You send a function code 90 on the network, and it will tell every other controller on the network to basically disconnect, so you can shut down the entire control network with one packet, based on this protocol.
I don't want to throw the protocol because they're another OEM competitor of ours. I'm being graceful. I'm just saying that stuff was built back in the day that we've inherited that now we have to refresh and really think with our security lens first because stuff that was built out of … productivity enhancement was never really thought. Would someone leverage this for maliciousness later on? Because it was it was like, hey how do we do this and how do we connect and how do we make our lives easier, and let's make higher efficiency, higher productivity. And in doing that we sometimes skip.
And that's some of the fears around AI these days is that people are taking and skipping some of the security guardrails that should be really thinking about before. Because in 10 years, what does that look like? We're inheriting with all the other systems that we have now that we're trying to secure or backwards secure.
MC: I think sometimes when cases like that, it's just that you can't look into a crystal ball and see the future. And if you could, a lot of our industries would be different.
RK: Thanks for having us today. My name is Rick Kaun. I practice cybersecurity in the OT realm. I have done so for about 26 years now. I've been all over the world with that job, it's really quite a fortunate opportunity for me to learn and meet people, but also to pick up on some trends and maybe some mistakes that some have made, and help to accelerate other people's journeys. So, I currently occupy a global director role at Rockwell Automation, which really allows us to get into lots of industries and lots of geographies, and again, even further enhance that database of where we're seeing what works and what people are struggling with, and how to maybe try to overcome some of that stuff.
MC: And I'm sure Rockwell has done a lot of research into these topics. Specifically, today we're talking about a report called Operational Resilience in the Age of Connectivity. Do you have any data you can present to us to start out about how cybersecurity impacts downtime in manufacturing?
RK: Yeah, there's lots. I mean, there's no shortage. I mean, obviously, our own report looks towards general trends around how many think they're ready, what they think they're exposed, where they're at. I think we've seen here that, for example, 37% say securing IT/OT architecture will drive positive business outcomes, right? Which is a great plug towards why you should be doing these things. But up to 46% is one of the things we found that organizations experienced a cyber incident the past year.
The thing that follows from that, though, is an incident is an incident, and it's always measured in downtime. Are we recovering? Great. I would think the next turn of that wheel would be: Are we recovering faster, and are we having less of them?
Today we're seeing that people are realizing what's there. They're recognizing it. They're continuing to invest in this, and they are continuing to make progress, which is all positives. If you're a glass half-empty person, there's room to improve. There's lots more we could do, and that's where we want to keep always highlighting this topic. Again, why get so excited to have opportunities like this. We're doing great, but we can't rest. Right, the bad guys aren't. AI is evolving. Risk is evolving. We need to make sure that we stay in front all the time. It's an always-on thing.
MC: I think it's topical that you bring up AI. That's something that I wanted to ask a little later, but I'm going to ask it now instead because you brought that up. Of course, AI right now is being used a lot by both threat actors and by cybersecurity teams, by those defenders, as well as by people every day. You might just be working on something in a shop and talk to ChatGPT and say 'what do I do here?' So, how has the use of AI changed cybersecurity considerations in industrial and manufacturing settings?
RK: Yeah, that's the question everybody wants to know. So let me break it down into a couple different scenarios. So first and foremost, AI for the bad guys is fantastic because they can use it to grab a piece of software or hardware and and hammer on it all day long with all this intelligence and horsepower that makes their job a little bit easier and more automated. Whether that's right or not, we can argue about AI slop and accuracy some other day. But the general premise is true that it's a great benefit to the bad guys. So what are the good guys doing? Well, they're doing the same thing. We can test our own stuff. We can look at it – we always do anyway, but much more thorough. What are they seeing? Can we see at the same time or faster? And then using that to get into then the proactive side of things.
So there's a straight 20,000-foot view of “AI is no good for defense”. Well, that's not entirely correct. We could be much more proactive, and we are getting much more diligent in using AI to understand where the holes and the cracks in the armor may be and get them protected as soon as possible. But when you take AI to the next step, which is “I now want to be proactive with this stuff.”
You can look at any sort of antiquated environment, and there's some industries that are … I don't want to say worse, but are more complex and more legacy than others, just in the way that they're built in the set and forget frame framework. That is most core building like power and oil. I mean, refineries are complex, but some of the core stuff is 40 years old, 50 years old. And what you could do in those is take all that complexity of all the assets and all the updates and upgrades you may need, and some of those, pre-plug-and-play PLCs and controllers. And how do you actually make heads or tails of first, where do I need to spend the precious time and resources of all the risk I have, I probably can identify 20,000 different things, literally vulnerabilities, obsolescence, many configurations? How do I prioritize? Well, AI can help me with that. And then once I want to prioritize, how do I then actually build a rollout that doesn't have maximum downtime?
MC: And I think another key here is asset visibility, making sure you can see those legacy systems, those temporary connections. Anything you want to add about making sure asset visibility is as beneficial as it can possibly be to maybe remediate existing security problems or get ahead of problems that may exist in the future?
RK: Visibility is king, but it's only one component. The most success in the rapid acceleration of maturity and shift from reactive to proactive is coming from organizations that provide context to the data. I can give you a list of of IPs and MAC addresses. We actually recommend about five different dynamics for what you want to build around an asset, and only one of the dynamics is IP, MAC address, vendor, what is it, etc. We want to know the context of the asset. Is it a critical safety system? Is it a redundant system? Is it near the DMZ or the business where it can more easily get compromised? Is it obsolete? Is it redundant? Is it backed up? We want to know all of those things, because they all have a factor in understanding the impact potentially to operations, and so visibility is not only key; it is the foundation. The asset must be your record, of the instance of the record, the base of your record, or your CMDB, if you will. And then onto that you build, okay, it's an asset, but here's the criticality.
We had a global mining client get told from a very well-intentioned IT team that said, "Hey, there's a critical flaw. Go do something about it." But when they looked at it and were preparing for shutdown, they realized they were they were secondary devices they could pull offline, get into lab, and patch there, as opposed to taking a full outage. The context was huge to avoiding that. Now, now multiply that against critical risk on spinning equipment or safety systems, right? Like that was sort of a trivial, “Hey, it's not a big deal. We can do it over here.” But imagine if you had to do that at scale, hundreds and thousands of vulnerabilities across hundreds and thousands of assets. You need to sort through that, and it comes from context.
MC: Cost is always part of that conversation too, and with maintenance and reliability in particular, sometimes it can be difficult to argue for solving problems that haven't yet occurred. That’s a saying that maintenance and reliability is working when no one notices that it's working, because nothing happened. So, any advice you would give professionals who are looking to argue for better cybersecurity oversight? “We haven't already been breached. We want to protect ourselves in the future.” What ROI is that going to have?
RK: Great question. And it's why I keep talking about cyber as being part of resilience and uptime, and not just about cyber for cyber sake. Because to your point, if you do everything a cyber guy tells you to do, nothing will happen. Where's your calculation, right? That's an oversimplification. Nobody's bulletproof.
But one of the things about context, which is absolutely key to addressing exactly what you're talking about, is we can look at an asset as the base record, and then when I talk about the impact of it, we have clients and organizations that are adding revenue generation by location or by site, or regulatory potential impact by asset or production line or facility. We have other ones where they have explosive chemicals and products at certain stages, but not in others, and so they have what they call the boom factor as part of their risk prioritization. And then what you're able to do is take the fact that we're seeing a percentage of our assets that are obsolete or not secured or not backed up, and it's at my number one revenue generator.
Well, now we can start to have somewhat of an empirical discussion, even if it's not predicting, well, you were going to have eight incidents, and we're only going to have four if you do this, because you can't tell the future. But what we can look at is how much risk have we accrued? What's our technical debt? And on which assets and which production lines, and relative to the corporate standards around safety revenue regulation, now we're having a completely different discussion. Now it's very empirical around things that the corporation reports on and the shareholders expect of us, right? So we're going from speculation to empirical data, and we can show that if we did this … We had a client that showed $2 million upgrade of obsolete PLCs would take a lot of risk out of a place that in total generates over a billion dollars’ worth of income for that organization became a no brainer.
MC: If you could give anywhere from one to five pieces of practical advice for maintenance supervisors and reliability engineers in terms of keeping their digital assets secure, in terms of increasing operational resilience with OT cybersecurity, what would your advice be?
RK: The number one would be obviously get a hold of everything that you have. There's two dimensions. I jumped immediately into the multi-dimensional view of the asset, not just a singular view, but also the first is that I know all assets, right? Some people have nested PLCs, devices, or they may have a cellular modem and a hub-and-spoke scenario in the field for transmission, distribution, pipelines, etc. You need to make sure you have everything. And I'm not trying to boil the ocean and say everything has to be locked down 17 ways from Sunday, but you need to know how many you have of what. So, if something comes up, you're prepared and you can pivot. So that's the first piece of advice: know your know your assets.
Number two would be to make sure that you capture the tribal knowledge. When I talk about asset context from an operational perspective, a lot of that is in people's heads, in people that are aging and potentially leaving the workforce, or not, being in the same organization. It's a struggle to get young talent to come out and work at a coal-fired power plant, for example, or a refiner. Or you drive to beautiful Sydney, Australia, and you drive out to the countryside to go to an aluminum smelter. Like, how do we find people to do these maintenance jobs and these work in these in these rural places? We need to capture tribal knowledge, and part of that goes into the record that I'm talking about building for assets, that allows you to then pivot into well, now we need to do security. Oh no, okay, it's secure, but what's next?
You're going to be asked for autonomous operations or for future industrial improvements, and so, do my controllers and my capabilities and my assets live up to high-speed data sampling and sending stuff outbound to be able to do digital enhancement and product improvements, et cetera? So, if you're the maintenance and the operations people in these facilities, these devices no longer sit in their own world, siloed behind a firewall that nobody ever cares about, everybody cares about them and what's coming out of them. So this is your chance to get in front of what do we have, what does it do, and how's it going to serve that higher purpose? That's where, especially if you understand that higher purpose, where it becomes easier to speak in the terms of the people who sign the checks to let you upgrade those things, or enhance them, or make them redundant. Right? It all comes from evidence and data.
MC: All right, those are all of my formal questions. Is there anything else you'd like to add, or that you feel we didn't cover?
RK: I just want to double down a little bit on the IT/OT convergence. I mean, there has been for the longest time, sort of a divide between the two, and in many cases, I … what do you call it? My wife and my headstrong 20-year-old daughter get into an argument, and I'm advocating both ways, right? And it looks like I'm talking out both sides of my mouth, but I'm not. I want to tell the OT people we need to trust and use the IT people when and where it makes sense to help us understand risk, because there's more of it likely than we're necessarily aware of. But I really want to implore the IT people to say, "Look, we know what you're trying to do. We know it looks like IT over here in the in the OSs and the switchgear, but there's a lot of stuff. There's that whole that third rail PLCs, relays, controllers, safety systems. You don't know anything about. You absolutely have to be at the table together."
And then the second would be to not get into fear, uncertainty, doubt, and speculation, but like practical understanding of what we have ... almost 50% of our respondents in our state of manufacturing report plan to apply AI and machine learning to cybersecurity. Which means we're going to be in these environments getting more data. We have to be prepared for it. And as the maintenance team and the and the and the front line staff that you're primarily talking to, and that we're talking here to here today, we can we can embrace that as an invasion of our space, or we can embrace it as here's what we're going to serve up and why it's important, and use that opportunity to educate and maybe increase our lot in life in terms of support, understanding, and maybe some more equipment or some more people. Capture the tribal knowledge. Be proactive in your upgrades. Be more redundant if when and where.
About the Podcast
Great Question: A Manufacturing Podcast offers news and information for the people who make, store and move things and those who manage and maintain the facilities where that work gets done. Manufacturers from chemical producers to automakers to machine shops can listen for critical insights into the technologies, economic conditions and best practices that can influence how to best run facilities to reach operational excellence.
Listen to another episode and subscribe on your favorite podcast app
About the Author
Megan Crouse
Megan Crouse has a decade of experience in business-to-business news and feature writing, including as first a writer and then the editor of Manufacturing.net. Her news and feature stories have appeared in Military & Aerospace Electronics, Fierce Wireless, TechRepublic, and eWeek. She copyedited cybersecurity news and features at Security Intelligence. She holds a degree in English Literature and minored in Creative Writing at Fairleigh Dickinson University.




