Skip to main content
    All Episodes
    Episode 093 · November 19, 2026 · 24m listen

    How to Measure Exploitability in Medical Device Cybersecurity | Ep 93

    Key Takeaways

    Cybersecurity risk assessments must measure exploitability rather than probability because cyber...

    Cybersecurity risk assessments must measure exploitability rather than probability because cyber attacks involve intentional, motivated adversaries rather than random statistical hardware failures.

    STRIDE functions as a threat classification model during threat modeling

    whereas CVSS 4.0 base metrics provide an objective framework for scoring exploitability.

    Medical device cybersecurity aligns AAMI SW96 security impacts with ISO 14971 patient harms

    primarily focusing on how integrity and availability compromises affect clinical outcomes.

    Relying on physical access constraints does not eliminate cybersecurity risk

    and the FDA frequently issues submission deficiencies to manufacturers who fail to implement controls on physically accessible devices.

    Penetration testing acts as an objective

    non-probabilistic validation method to confirm whether technical controls successfully reduce device exploitability.

    Episode Summary

    In this episode of the Med Device Cyber Podcast, host Christian Espinosa breaks down why medical device manufacturers must move away from probability and likelihood when evaluating cybersecurity risk. While probabilistic models work well for standard mechanical wear and tear under traditional risk management frameworks, cybersecurity involves intentional, motivated adversaries rather than random statistical events. Consequently, regulatory bodies like the FDA expect medical device risk assessments to evaluate exploitability rather than probability. Christian explains how product security teams and regulatory leads can bridge standard security risk frameworks like AAMI SW96 with ISO 14971 patient safety requirements. The discussion outlines how STRIDE should be utilized as a threat classification model rather than a scoring methodology, and how CVSS 4.0 base metrics—specifically attack vector, attack complexity, attack requirements, privileges required, and user interaction—provide an objective measurement of exploitability. Finally, the episode tackles common submission pitfalls, warning teams against assuming physical access constraints eliminate cybersecurity risk. Because hospital rooms represent hostile, accessible environments, manufacturers must still design layered controls and validate them via penetration testing to drive exploitability as low as possible.

    Chapters

    1. 0:07Hi, welcome back to another episode of the Med Device Cyber Podcast
    2. 3:16use probability. They, if they're motivated, have an intent, they look at
    3. 6:24exploitability. So a little bit more about exploitability in terms of how
    4. 9:32difficult. And this is the main factor. And if you just think
    5. 12:40known as a CIA triad
    6. 15:48looked at it in terms of cybersecurity
    7. 18:57devices are in proximity to a hospital bed that are accessible
    8. 22:05do. We can reduce it. We can't get rid of it. But

    Frequently Asked Questions

    Quick answers drawn from this episode.

    • In this episode of the Med Device Cyber Podcast, host Christian Espinosa breaks down why medical device manufacturers must move away from probability and likelihood when evaluating cybersecurity risk.

    • Cybersecurity risk assessments must measure exploitability rather than probability because cyber attacks involve intentional, motivated adversaries rather than random statistical hardware failures. STRIDE functions as a threat classification model during threat modeling, whereas CVSS 4.0 base metrics provide an objective framework for scoring...

    • This episode covers Threat Modeling, Penetration Testing, and FDA Premarket Cybersecurity. It's part of The Med Device Cyber Podcast, hosted by Blue Goat Cyber, focused on practical medical device cybersecurity guidance for MedTech teams.

    • Consequently, regulatory bodies like the FDA expect medical device risk assessments to evaluate exploitability rather than probability. It's most useful for medical device manufacturers, cybersecurity engineers, regulatory affairs professionals, and MedTech founders preparing for FDA review.

    • Cybersecurity risk assessments must measure exploitability rather than probability because cyber attacks involve intentional, motivated adversaries rather than random statistical hardware failures.

    Listeners also asked

    Quick answers pulled from related episodes.

    • What does Episode 25 cover about "How to Make Your MedTech Startup Investable with Gavin Leonard"?

      What actually makes a MedTech startup investable? Is it the technology, the clinical need, the founder, the pitch, or whether the company has simply done enough homework before walking into the room? Christian Espinosa speaks with Gavin Leonard of MedTech Syndicates about what...

      From Episode 025 · How to Make Your MedTech Startup Investable with Gavin Leonard | Ep 92
    • What does Episode 24 cover about "Why Medical Device Penetration Testing Is Different"?

      What makes medical device penetration testing different from a traditional cybersecurity penetration test? In this solo episode, Christian Espinosa explains why medical device penetration testing has to be approached through the lens of patient safety rather than simply data...

      From Episode 024 · Why Medical Device Penetration Testing Is Different | Ep 91
    • What does Episode 23 cover about "Why Medical Device Companies Need Tabletop Exercises with Daniel Hammond"?

      What happens when a medical device company experiences a serious cyber incident? Daniel Hammond joins Christian Espinosa to discuss why incident-response decisions should be made before an organization is under attack. They explore the role of tabletop exercises, why ransomware...

      From Episode 023 · Why Medical Device Companies Need Tabletop Exercises with Daniel Hammond | Ep 90

    Share this episode

    Pre-fills with: "Cybersecurity risk assessments must measure exploitability rather than probability because cyber attacks involve intentional, motivated adversaries rather than random statistical hardware failures."

    Show original episode notes
    How do you calculate cybersecurity risk when the thing you are measuring has an attacker making decisions on the other side? Probability works when there is historical data to lean on. Elevator cables wear out. Steel fatigues. Tornadoes follow patterns. A motivated attacker does not politely wait for the statistics to catch up. This week, Christian Espinosa makes the case for exploitability instead: how difficult is the attack, what access does it require, and what stands between an attacker and patient harm? That distinction matters even when a device requires physical access, because a USB port beside a hospital bed is only reassuring until somebody decides to plug something into it. Parachute pants had their moment. Probability-based cyber risk may have had its moment too. In this episode: 00:07 Why exploitability replaced probability 02:19 Why probability needs historical data 04:30 The condo analogy: likelihood versus exploitability 06:33 Using STRIDE to classify medical device threats 07:37 How CVSS 4.0 measures exploitability 12:06 Connecting cybersecurity risk to patient harm 14:19 Why penetration testing validates exploitability 17:42 Why physical access does not remove cyber risk 20:38 Why low risk still needs mitigation 22:42 Designing devices for lower exploitability The Med Device Cyber Podcast is brought to you by Blue Goat Cyber, cybersecurity experts providing essential security solutions for the medical device industry. Learn more: https://bluegoatcyber.com Schedule a Discovery Session: https://go.bluegoatcyber.com/meetings/blue-goat-cyber/discovery-session Christian Espinosa on LinkedIn: https://www.linkedin.com/in/christianespinosa/ Blue Goat Cyber on LinkedIn: https://www.linkedin.com/company/blue-goat-cyber/ Blue Goat Cyber on Instagram: https://www.instagram.com/bluegoatcyber/ Blue Goat Cyber on Facebook: https://www.facebook.com/bluegoatcyber/ Blue Goat Cyber on YouTube: https://www.youtube.com/@BlueGoatCyber/?sub_confirmation=1 Download MP3 — full interview
    Hi, welcome back to another episode of the Med Device Cyber Podcast. I'm your host, Christian Espinoza with Blue Coat Cyber, and today we're talking about exploitability, which is a critical thing to discuss because we have migrated away from probability and likelihood when we calculate risk for medical devices because we've realized probability and likelihood are no longer valid. They may have served a purpose, kind of like parachute pants in the '80s. It was a trend, but now the trend, which I agree with, is exploitability. So imagine you've got a cardiac monitor, it's monitoring somebody you love, it's on their nightstand at night, and we ask the question, how likely is an attack tonight? That's how we used to do things. That's probability or likelihood. Instead, we really need to be asking, how hard would it be for an attacker to pull off the attack? That's exploitability, and that's what we're focused on today. It's a more objective means of calculating risk when we look at one of the axes of risk. So today we'll talk about why probability breaks, why we should be using exploitability, STRIDE, which is a way to classify threats, CVSS, ISO 14971, and SW96. I know that sounds like a lot of acronyms and a lot of, like, babble talk or robot talk, but I'm just gonna put it— frame it in context of exploitability. So with ISO 14971, this is extremely important with medical devices. This looks at severity of harm to a patient. So what effect can I cause to a patient, and what is the harm to a patient, which is ultimately all that matters with medical device cybersecurity. So we look at the risk as a severity of the harm to the patient combined with probability, which is what we used to do. Now we're combining the severity of harm to a patient with exploitability in terms of risk that are related to cybersecurity. This makes it more concrete. An example of probability, and probability has a use, and it still has a use today, but not in terms of hacking or cybersecurity. Probability is typically based on statistics, like what is the likelihood of a tornado coming and hitting Clarksville, Arkansas, where I grew up? Well, it's fairly likely, one every 5 years, probably. That's why we had a tornado shelter going up, uh, and everyone would go hide in the tornado shelters when the tornado sirens went off. That's why there's tornado sirens as well. So that is based on probability. Another example is an elevator cable. There's a cable that pulls the car up and down elevators. After a while, that cable starts wearing out after so many up and downs in the elevator. So when is that cable going to break? Again, that's probability or likelihood. Probability and likelihood requires a sample of historical data. We typically don't have that with a cyber attack. Which is why we want to use exploitability. Attackers don't use probability. They, if they're motivated, have an intent, they look at how easy it is to accomplish something. So probability is no longer the right way to look at calculating risk in terms of medical device cybersecurity. And with Amy SW96, which is just a standard we follow for security risk With Medical Devices, it talks about the challenge of using probability. So instead of using probability, we want to use exploitability. Probability is fairly subjective and it requires a lot of historical data, like I mentioned. So we wanna look at intentional attacks, which is cybersecurity versus accidental faults. It's fine for accidental faults to be classified in terms of probability, like the elevator, like the steel on a bridge over time may fatigue and the bridge may fall down. So we, we, that's accidental fault. We, we know the data on that. We have the data. That's very different than somebody with intentionality going to the bridge with a torch and melting the steel joints to hold the bridge together. That is intention, and that's like a cybersecurity attack. Security is intentional attacks, which is why exploitability should be used. Probability is irrelevant. It's kind of like if I look at my condo here, I could say based on probability, since I've lived here, which is almost 2 years, nobody has broken into it. Does that mean I should not put locks on my door or I should leave the doors unlocked or I should leave the windows unlocked? I should turn the cameras off. I should just assume no one's ever going to break into it because over my historical timeframe, I've, since I've been here, the probability would be zero at this point. But what is the exploitability? That's what I'm more comfortable with. If someone were to try to break in, do I have locks on the door? Are those locks sufficient? How hard am I making it for the criminal to break into my condo? Same concept with that cardiac monitor at night. How hard is it for someone to break into it? Because if they have intention and a motive, the probability doesn't matter. We have to look at how difficult or hard it is to take advantage of that vulnerability, which is the exploitability factor. The FDA wants us to look at exploitability. They don't want us to calculate risk based on probability anymore. So we're really looking at how hard is it to turn a weakness into a real effect. That's the exploitability. So if we've got doors, one with no lock, one with a deadbolt, one that's a vault, which one is more easy to exploit? Obviously the one with no lock. So that is how we calculate the risk, exploitability. Security risk should be non-exploitable. Probabilistic, which means that we're focusing on exploitability. So a little bit more about exploitability in terms of how we classify threats. One of the things we use is STRIDE. So when we're looking at ways to exploit a medical device, we have to look at a threat model. A threat model is all the ways somebody can attack that device. And then we look at what they can do if they're able to carry off the attack. And that's where STRIDE comes into play. And then we look at ultimately how can we reduce the exploitability. So STRIDE stands for spoofing, tampering, repudiation, information disclosure, denial of service, and elevation of privileges. These are what we could do if we're able to realize a threat or exploit a device. Can we spoof the data on the device? Can we modify the integrity, which is tampering? Can we gather sensitive data or proprietary data, which is information disclosure? So this is just a way of looking at the threats and what we could do if we are able to realize that threat or break into that entry point. And a lot of people can view STRIDE, they think it's a scoring mechanism. It's not, it's a classification mechanism. It's a way to classify things we can do when we're able to realize a threat. For actually scoring the exploitability, we use CVSS 4.0. There's also CVSS 3.1, but 4.0 is a good mechanism for looking at exploitability. And really all you need to look at are what's called the base metrics. The base metrics are really how hard is it for someone to take advantage of a vulnerability or to exploit this medical device. One of them is the attack vector. This is the one that drives the score the most from low, medium, or high because we look at, do I have to have physical access to exploit this device? So physical, local, adjacent, or network are the 4 categories. Network is on one extreme. That means I could sit inside Starbucks, I could be at my condo, I could be driving down the road connected to my cellular hotspot. And try to attack the device because it is software as a medical device on the cloud, as an example. That is highly exploitable because I can try to exploit it from anywhere, all the time, 24/7. In contrast, if the attack vector requires me to have physical access, I have to physically be at the device that makes it very difficult to exploit. So this will be low exploitability because now I can't do it from Starbucks. I can't do it from my car, the airplane, anywhere Wi-Fi. I have to physically drive to the facility or get my hands on the device in order to try to exploit it. So this makes the exploitability very difficult. And this is the main factor. And if you just think about it, it makes sense, right? If I have to physically be in a hospital or physically have the device or somebody has to physically break into my house to connect to that cardiac monitor because it only has a USB port, it's much more difficult to exploit than if that cardiac monitor has Wi-Fi and Bluetooth. Now, if it has Wi-Fi, it may be on the internet. So Anybody can try to attack it all the time. I can look at Shodan to see if it's available up there. If it has Wi-Fi or Bluetooth, if I'm in a condo or an apartment, I might be able to see it and try to remotely attack it if it's, you know, one of my neighbors or someone upstairs or downstairs. If it's in proximity where I can see it, then I can attack it. So that makes it much easier than if it only has a USB port. The second base metrics in CVSS is attack complexity. So how hard is it to pull off the attack? Do I need to have a large skill set to pull off the attack or is it fairly easy? The third one are the attack requirements. Are there requirements for me to carry off the attack? Do I need to intercept the traffic, for instance, and then decrypt the traffic? Does the device have to be in a very specific state? The next one are privileges required. Can I just attack it with no account at all? Or do I need to have a user account or an administrator account on that device? So you see, when we go through the exploitability metrics here, they all play a role. The one that matters the most is proximity, basically the attack vector, but the other ones play a role as well. So if I have to attack it remotely, so it's much more easy to exploit because I can do it from wherever, but the attack has to be very complex. I have to be able to do a lot of things from ATT&CK requirements. I need privileges. Then the exploitability can go down, even though it's accessible across the internet. The fifth one is user interaction. Do I need to trick a user into doing something? Again, this affects the exploitability. If I can carry off the ATT&CK without tricking the user to do anything, then it's obviously more exploitable. But if I have to trick the user to click on a button or to click on an email, or to open a file, then exploitability is more difficult because now I'm dependent on the user. So as you can see, we're taking this in a more granular approach than just saying, you know, the odds are 1 in 10,000. Now we've made it more objective, specifically in the terms of what the attacker or the malicious software that's propagating the internet must be able to accomplish for the attack to succeed. This is the exploitability. And the way we look at this at Blue Goat Cyber is we follow the standards, Amy, SW96, which is really security risk. So how do we classify risk from a security perspective, which we typically look at confidentiality, integrity, and availability, also known as a CIA triad. So we look at the exploitability of a device and what the effect could be in terms of CIA. And we also apply STRIDE. But then if one of those effects that we can cause can cause harm to a patient, we have to look at the severity of harm to a patient, which is ISO 14971. So we take our risk from AMES W96, the security risk, which is typically Confidentiality, can I modify, can I steal sensitive data? That's confidentiality. Integrity, can I modify how the device works or switch the diagnosis from positive to negative or vice versa? And then availability, can I take the device offline or make it a delay? Typically the two things that affect a patient are integrity and availability. Integrity. Can I alter the diagnosis? Can I— diagnosis. Can I say there's a true diagnosis when there's a false one, vice versa? Can I create a misdiagnosis? Somebody has sepsis, can I say they don't? If somebody has cancer, can I say they don't have cancer and vice versa? And what is the effect of that to the patient? And then availability. Can I cause a delay in treatment to the patient? And what is the effect of that delay? So then we have to look at how that can affect a patient, where— that's where ISO 14971 comes into play. And really we're looking at what's the threat, what's the control, what's the test to validate the control, and what is the effect to patient harm. And the test is how exploitable it is. Effectively, a penetration test is not probabilistic. We're going to try to attack the device at that point in time, just like an attacker. We have the intent. So we need to validate that the control that's supposed to mitigate the threat works. This is like my condo. I have a door over there. I've got a lock on it. I don't know if that lock is actually gonna work. So we wanna test the lock. We wanna see if I can just put a key in there and get it to unlock or easily, et cetera. I do have a deadbolt lock, so it's probably unlikely to pick that, but still something you wanna look at from exploitability perspective. And if you think about in your daily lives, you probably think in terms of exploitability, you don't even realize it, uh, instead of probability. So if you go to a parking lot, you may want to lock your car, you may want to park under a light if it's dark, uh, and those are things that you are looking at. Well, if it's well lit and my doors are locked, and back in the day you could put that club thing on your steering wheel, that I'm making my car less exploitable for a burglar than if I didn't park under the light, if I left it unlocked or the windows down, et cetera. And you probably didn't think, when was the last burglary from a probability perspective in this parking lot? If you thought that and there were no burglaries, you probably wouldn't even lock the car. But we look at exploitability on a daily basis. We just haven't really looked at it in terms of cybersecurity. So now we need to look at in terms of cybersecurity. And I think the overall objective here is to make our device very difficult to exploit. So a criminal, a nation-state, malicious software can't exploit our device very easily. So our exploitability should be difficult, which will overall lower our risk. And one way to make this difficult, obviously we talked about before, is to require physical access. But that's not very feasible today, especially since we're migrating to connected devices, connecting our medical device with the EHR or EMR or a PACS. That requires connectivity, which gets rid of the requirement for physical access. So now we have to look at what else can we do from a control perspective to reduce the exploitability, because exploitability is what matters the most in terms of risk. So we can, can we make it extremely difficult for someone to attack our device, which is increasing that attack complexity? Can we add a lock on the door? Can we have two locks on the door? Can we have a siren that goes off? Can we make sure it has to have a special key or special combination as example? This makes the attack more difficult, which decreases the exploitability, which is what we want, low exploitability. What else can we do? Can we require privileges? So somebody has to have some sort of authentication on the device before they can do something. Can we increase the attack requirements? Again, all these things make it more difficult to exploit the device. And this is what we really want to look at. And these should be designed into the device, which is why a threat model should be done early on. You should identify the threats while you're doing development and while you're doing design, actually, and look at what design controls you're going to put to mitigate those threats. And then if we look at exploitability, we're going to look at, oh, okay, this needs to be very difficult for someone to pull off an attack or this threat. So what are we going to do to make that difficult so we can get the exploitability down to a low factor? It's not very exploitable. And a lot of people will say, well, my device requires physical access, so I don't need to do anything because by default it's low exploitability. Not true. Just because you have low exploitability doesn't mean you don't need to do anything. The assumption is someone is going to get physical access. And the FDA talks about this, and we see many of our clients that have received deficiencies because they assume because someone required physical access, they didn't need to do anything. Not true. You still need to secure the device. Even if this device is in the operating theater in the hospital, or it's an MRI machine, we consider those environments hostile environments. I've been to hospitals and visited people in hospitals. I've been a patient in a hospital when I had blood clots, and I was surprised by how many devices are in proximity to a hospital bed that are accessible. To pretty much everybody that walks in and out of the hospital. On average, I think I heard there are f— well, I did hear there are 14 medical devices connected to a hospital bed. That's 14. 14 that probably have vulnerabilities. 14 that have probably been there a long time, that are legacy devices. 14 that, out of those 14, a few probably require physical access. A lot of them probably have Bluetooth. A lot of them probably have Wi-Fi. Some of them may have an Ethernet connection. And these are all devices that can be compromised. So we need to make it very difficult to exploit these devices because at the end of the day, we're talking about patient safety. And when I visited my cousin in Denver a couple years ago, I was looking at all the devices around the hospital bed and I saw USB ports. I could see Bluetooth on my phone. I could see lots of Wi-Fi connections. You know, I didn't try to hack into any of them, but I thought, I bet some of these are very vulnerable. And there was nobody monitoring me when I was visiting my cousin. Nothing to stop me from putting in a malicious thumb drive, plugging in a keyboard, or a Bash Bunny, or any device into the USB port on some of these medical devices. There was nothing to stop me. So even though physical access is required, we still need to do things to reduce the exploitability, which means might— we might need to white label what types of devices can be plugged into the USB port. We might need to cover the USB port with tamper-proof seals or whatever the mechanism is. And I'm not to belabor a point, but this has happened more times than not. We've had clients argue they don't need to do any cybersecurity because their exploitability is low because of physical access. But the FDA has kicked back their submissions and said, that doesn't mean you don't need to do anything. You still need to do things because ultimately there's still a risk. The risk may be low, but there's still a risk. So we still have to do things to mitigate the risk. Ultimately, at the end of the day, cybersecurity is about risk. And the only way to remove all risk is to not have that medical device deployed. Just like when I used to skydive, people would say it's a risky activity. And I would argue, well, it's not that risky if you take the steps to reduce the risk. Uh, like you are trained for emergencies, you know what to do if your parachute doesn't deploy, if it deploys wrong, if your lines are twisted, you have an automatic activation device. So if you're knocked out or something happens, it'll automatically deploy your reserve parachute. If you're going a certain speed at a certain altitude, uh, you wear a helmet, you have two altimeters, you only jump with the right people at the right drop zone. There's a lot of things you could do to reduce the risk, but the only way to get rid of all the risk is to simply not skydive. And it's the same with medical devices. We need to do the things to reduce the risk or reduce that exploitability because that's all we can do. We can reduce it. We can't get rid of it. But just because it's low doesn't mean we shouldn't do anything about it. So to sum up, I know I've kind of talked a little bit about CVSS quite a bit and exploitability, but to sum up, we need to focus today on exploitability as one of the axes of our risk matrix, not probability, not likelihood. Probability and likelihood have a place for things like failure rates, for things like the cable in the elevator, for things like the strength of the steel on a bridge over time, for things like the weather, that has a place. But when it comes to attacks, cybersecurity attacks or cyber criminals, we need to look at how easy is it for them to carry off the attack. Just like the scenario I gave with your car in the parking lot. I'm pretty sure none of you look at probability. You probably lock your door anyway, and you're going to park under the light at night no matter what. Some of you may say, I live in a safe neighborhood and choose to not do that until somebody breaks into your car. And then after that, you'll probably lock your car everywhere. Again, we're looking at intentionality with attackers. So exploitability, CVSS 4.0 is a way to measure exploitability. It's objective. The 4 metrics I— or I mean the base metrics and the 5 base metrics I talked about. Are important. And then we're just designing our device. We look at how can we reduce the exploitability. And this includes things like, can it be physical? Is physical access required? Is remote access required? If you have a CMD, it's going to be remote. So automatically it's going to be difficult. It's going to be easier to exploit. So I can affect the attack complexity. What requirements are required? Are privileges required? Does it require user interaction? There are things I can design into my device to make it more difficult for someone to exploit. And then since that will change exploitability, that will ultimately reduce the risk to patient harm, which is really what we're focused on. A takeaway for me, I used to think in terms of probability. In fact, in my first book, "The Smartest Person in the Room," which I published quite some time ago, I wrote about risk in cybersecurity in terms of probability times impact. And that was pretty valid back then, but since then we have migrated to exploitability. And now if I make a new edition of that book, I will change the matrix to focus on exploitability. Things evolve, attacks evolve, and we need to evolve as well. So thanks for tuning in. I look forward to seeing you in the next episode. Take care.

    Hosted by

    Explore every episode in the topics covered here.

    More from your host

    Other episodes diving into Christian's areas of focus.

    Episodes covering similar ground - including Threat Modeling, Pen Testing, FDA Premarket.

    Why this matches shares the Threat Modeling, Pen Testing, and FDA Premarket topics.

    Why this matches shares the Threat Modeling, Pen Testing, and FDA Premarket topics.

    Why this matches shares the Threat Modeling, Pen Testing, and FDA Premarket topics.

    Why this matches shares the Threat Modeling, Pen Testing, and FDA Premarket topics.

    Listen to this episode