Summary
Michele Mosca, founder, President, and CEO of EvolutionQ, a quantum-safe cybersecurity company is interviewed by Yuval Boger. They talk about Post-quantum cryptography and quantum key distributions, the timeline for hardening a company’s infrastructure, standards organizations, and much more.
Key takeaways
- Migrating to quantum-safe cryptography is not a software patch, it requires inventorying every cryptographic dependency across an organization, which most companies have never done
- Mosca's inequality frames the risk: if the time to migrate plus the shelf-life of sensitive data exceeds the time until a cryptographically relevant quantum computer exists, the data is already exposed
- Quantum key distribution and post-quantum cryptography solve different problems and are not interchangeable, so treating QKD as a full substitute for PQC misses the point
- Finalized NIST standards are just the starting line, testing, integration, and supply chain coordination across vendors is the part that actually stretches migration timelines
Transcript
Yuval Boger: Hello, Michele, and thanks for joining me today.
Michele Mosca: Hello, Yuval. It's great to be here.
Yuval: So who are you and what do you do?
Michele: I'm Michele Mosca. I'm a mathematician by training. I lead the quantum-safe cybersecurity company evolutionQ. I'm involved in a number of other activities building up the quantum and quantum-safe ecosystem, and I maintain an affiliation at the Institute for Quantum Computing at the University of Waterloo, which I helped found many years ago.
Yuval: So I think we all owe a big thanks to Professor Shor for his algorithm, because that raised awareness of quantum and the risks that quantum poses for cybersecurity. How does your company help address these risks, and in what timeframe?
Michele: Yeah. So our company started off several years ago helping with this, 'cause there are several stages to getting your systems to be resilient to quantum-enabled attacks. We need that to be in place in order to reap all the positive benefits of quantum computing that your company and other companies around the world are working toward helping everyone realize.
But for that to really happen without very serious collateral damage, to put it mildly, we must first get everyone, or at least the major digital platforms, to be sufficiently resilient to quantum attacks. But that's not a one-step process. So years ago we started off with the first stage, which is really awareness, right?
Then in the second stage, we pioneered what's known as a quantum risk assessment: helping people internalize what this really means to them. How would this impact us? What are the high-order terms? What are the most serious problems we need to start addressing, and when? And then the third phase is starting to actually do so: coming up with a roadmap, a plan, engaging with the vendor ecosystem, testing solutions, piloting things, and preparing.
It's your last dress rehearsal before the final stage, which is, "Now I've deployed quantum-safe solutions in my various digital platforms." The first two to three stages are professional services, and what we discovered as we were doing these professional services for people over many years is that there was a gap. The gap was the following.
The first layer of defense to protect against quantum-enabled attacks is what one might, at least conceptually, look at as a plug-in replacement, in quotes, for the existing mathematical algorithms that enable HTTPS, secure browsing, and all these other things.
That's a long and hard process that we've been actively engaged in for many years, and that's largely been a part of our services. In fact, I helped launch an open-source platform to make it easier for people to benchmark and test these solutions in their systems, so we can be ready for the day when we really, really need to deploy these things, because the attacks are really imminent or are actually happening. But we can't just go back to the status quo, right?
In the '90s, if public key cryptography was broken, it would've been a bad day, okay? That's an understatement, but it wouldn't have been an existential threat, right? The economy wouldn't have collapsed. There wasn't a massive safety issue associated with cryptography, but now there is, right?
Now it's not just IT, but also OT: cyber-physical systems and so on. And guess what? The dependency on digital platforms, and therefore on their cybersecurity, is going to keep increasing as we go to 5G, 6G, machine-to-machine, and so on.
The stakes just keep getting higher. So we can't just accept the chance that these new codes could be broken, and at least two of the codes under consideration ended up being broken. Maybe they can be fixed, maybe they can't, but if they're deployed in production systems and they're broken, you don't get to hit the pause button, right?
So there's a lot we need to do to make our systems more resilient. A lot of that includes more agility, making it faster to switch from one code to another, and putting in some extra layers. But there's also a role, especially for the most critical systems, for a key that isn't susceptible to mathematical cryptanalysis. That's where quantum cryptography and quantum communication networks come in, and they continue to improve. I say QKD is a solution that ages really well: as technology advances, the advantage really goes to the defender in this case, 'cause you can deploy better and better QKD systems.
What was missing when we looked at this was an easy, scalable, convincing way to take QKD from a point-to-point, very hardware-centered solution to something you could deploy across an enterprise without worrying about the details of the hardware.
So we developed a vendor-agnostic framework where you could interoperate many different vendors, and do it in a very scalable way. That's our product, called BaseJump. It's a quantum delivery network. We coined that term, 'cause the idea is that it's a framework, a tool for delivering quantum-safe keys securely, rapidly, and robustly to endpoints on demand, so they can have that extra layer of assurance against cryptanalysis of the more conventional methods.
Yuval: So if I'm a CISO, a chief information security officer, it sounds like the assessment of my risk would take two minutes, because you'd say, "Well, you're using systems today that could be broken with a quantum attack, so you need to do something quickly, at least for the systems that are most important to you."
How long does it take to harden an organization and secure some of its systems? Is that a multi-year process? Is there a Band-Aid that you're suggesting right now, and a better one in two years?
Michele: Yeah, great point, 'cause that's often the perception: well, what's there to assess?
The risk assessment looks at all the different ways. First of all, it takes time even to find… People don't have an inventory of what cryptography they're using. So it actually is a pretty sophisticated process at the end of the day. You have to see the vulnerabilities, figure out how they can actually be exploited, start ranking the vulnerabilities in terms of likelihood and impact, and then have a prioritized approach in terms of which systems you really need to protect against quantum-enabled attacks, and by when, right?
So you do have to have a prioritized list. You really want this system protected by this year and that system by that year, and then there are the nice-to-haves and the systems that will get secure when they get secure. But what are the high-order terms?
The second part of your question really is the critical point. It takes many years to migrate these things, so you do have to prioritize, and you do have to get started in time. How long it takes can vary. In a test system, you could do it in a matter of weeks, but in any real-world system, it takes years to migrate, for many different reasons.
For one, there are dependencies, right? Again, it really depends on the system, but in most real-world systems, it's not the case that one company is both the client and the server and all the users use whatever operating system they're told to, right? So it's complicated when you're trying to take a legacy platform and migrate to a very new tool with different requirements.
It can take anywhere from five to 15 years to properly migrate a system. Now, people might say, "Well, can't you just try harder and go faster?" A little bit, but there is a fundamental speed limit at which you can change something so complicated, with so many different stakeholders, in a secure way. You want it to be certifiable, and we don't even have the certification framework set up, right?
That alone will take years, right? You can't just have everyone saying, "Hey, we've quantum-proofed it, trust me." We need those frameworks to be set up. We're working on all these things in parallel at this point, because otherwise we won't be ready, or at least we likely won't be ready in time.
So yeah, that can take quite a number of years, and the essential thing is to get started. One last point I wanted to make on this: if you try to break that fundamental speed limit, what happens? There'll be tremendous costs that do not require quantum attackers, right? So even if the quantum attackers don't come after you, your systems will crash simply because you rushed the migration. You didn't give people time to do the upgrades they needed to do. You will make mistakes, right? Systems go down because of a software update that didn't go as planned. Imagine doing something as fundamental and ubiquitous as this in a rushed manner.
There are no cyber attackers even involved there. Those are just self-inflicted wounds, because you created a crisis that was avoidable. But there's also additional cyber risk, in that when you rush the deployment of any software, but especially cryptography and cybersecurity tools, you're going to make mistakes in the design and the implementation, and cyber attackers love that.
That's their bread and butter. It's not typically sophisticated cryptanalytic attacks; it's exploiting mistakes that have been made, user mistakes, and so on. So if we try to rush this migration, it's a wonderful day for cyber criminals. They're already kind of winning the war, so the last thing we need is to give them yet another huge advantage.
Yuval: How does this work end to end?
Right now I'm using a Mac, and let's say I'm connecting to my bank, so I have a regular encryption protocol in place. The bank may be connecting to some other bank, and maybe in a couple of years that will be a quantum-secure link. But do you think I'll get a quantum key on my Mac in a couple of years, or is it just the backbone that's going to be more secure?
Michele: In the short term, and even the medium term, we're really talking about the backbone, where QKD keys are delivered at the base level. But there are certainly enhancements we can make even at the last mile to take it up a notch or two beyond vanilla-flavored end-to-end post-quantum.
I'm a big believer in vanilla-flavored end-to-end post-quantum, because at some level, the best security is the security you use. Remember, the cheapest and easiest thing is to do nothing, and for much of our communications, we don't encrypt, we don't authenticate, and we make a risk decision.
But the next thing you want is the easiest non-trivial defense you can deploy as ubiquitously and robustly as possible, so people don't find it a nuisance and aren't working around it, right? Post-quantum cryptography plays a beautiful role there. It provides some great protection, and it's beautifully end to end, right?
And then you go one layer down and protect even more. The whole point of a quantum delivery network is to make this higher level of assurance, this extra layer of assurance, easier to deploy and scale. But indeed, as you say, if that's going to connect certain basic nodes or hubs, you still have that last mile of protection.
If it's just the last mile, then it's a much easier proposition to deploy some of these higher-assurance methods that don't require quantum.
Yuval: And as it relates to your company, to the extent you're willing to tell me, do you primarily sell services like the assessment, or does the bulk of revenue come from products, like the distribution network you mentioned?
Michele: We're going to be transitioning now, right? Up until now it's been largely services, and the services part will grow, but the revenue from BaseJump licensing will overtake that over the next small number of years.
Yuval: Recently
Michele: But we're going to keep both.
We're going to keep helping with the services, but it's the deployment of the solution that's going to become the bigger part of our company.
Yuval: Recently NIST, the National Institute of Standards and Technology, made an announcement about the four initial protocols that they want to use.
My question to you is, first, how close is it to a working standard? And second, does that become a global standard, or is similar work going to happen in the EU and other places before it becomes a worldwide secure network?
Michele: Yeah. Sorry, what was the first part of the question?
Yuval: How mature are these NIST
Michele: Yeah.
Yuval: standards? How close are they to being a real standard?
Michele: Well, they're going to be a real standard by around 2024, right? How ready are they for real-world showtime? I think we're in the Goldilocks zone. Some people say, "It's too soon, it needs more time." Other people say, "We needed this a few years ago."
There's no perfect time, 'cause everyone has a different risk tolerance and a different migration time. I think it's going to be pretty robust, but that's why, at the end of the day, you need years of deployment in real-world systems to further build up confidence in these systems.
So the sooner you start, the sooner you'll build up that confidence, right? Furthermore, there's going to be a next phase of this confidence building when quantum computers are actually here. We can battle-test the implementation now, right? Interoperability and a lot of those real-world issues.
We can battle-test resilience against classical attacks now, so again, the sooner we start, the sooner we can get that done and build up that confidence. But the question of confidence against quantum attacks will remain until we've had large-scale quantum computers for several years.
So what do we do in the meantime? Well, that's partly why we continue to advocate for continuing to deploy the battle-tested, hardened current methods for key agreement and signatures. Of course, once quantum computers are mainstream, those provide limited protection, right? But it's not going to be free to launch quantum attacks, so they still offer some protection; they're still making it a nuisance.
But that's why you want an additional layer of defense, such as what QKD can offer. The fantasy is that we have a code with a guarantee that it's secure, and it's cheap, and it's easy, but that's just not the way the real world's going to play out.
But we're still in a pretty good place, where we have codes we have pretty good confidence in, and we're going to continue to build up confidence. And we have mechanisms for mitigating the residual risk. So we just need to be proactive as a society, right? Internalize these risks and costs, and properly manage them.
In terms of whether every jurisdiction is going to have its own standards, we generally hope not, right? For much of the world, NIST has an open process, and we tend to adopt the NIST algorithms, right? But then there are other standards built on top of that, 'cause it's not like the NIST standards are completed and we're done.
There are financial standards on top of that, and standards for how you use the cryptography in all sorts of different systems. That's where ETSI, ISO, and other standards organizations come in. There are going to be risk management standards, and lots of other standards around the core cryptographic algorithms.
There, again, NIST is, I think, the foremost authoritative body. It's USA-based, but many of the algorithms have come out of Belgium, right? So it's definitely international in how it engages people. That said, the German government has proceeded to standardize some algorithms before NIST. Remember I said the Goldilocks zone, and that some people needed it sooner?
Well, the German government felt it needed some algorithms. It needed something; given its risk tolerance, it couldn't wait. So it's standardizing some algorithms. Now, these are not out of left field. These are algorithms that were part of the NIST process but aren't necessarily on the NIST finalist list. That doesn't mean they'll never be standardized by NIST; it just means that in the short term, the German government is standardizing them.
But it's all, I think, done in a collegial, collaborative way, where all of society benefits. Now, there are parts of the world that may not want to adopt the NIST standards. But again, I hope that as a global society, we get to a place where we can agree on a simple and secure way to send zeros and ones across the planet, and compete at higher levels in the stack.
China's pursuing its own process. The fundamental machinery is very similar, but it wouldn't be surprising if they develop their own standards, and then there's going to be a bit of extra work getting all these different platforms to work in these different jurisdictions.
But I'm not expecting a plethora of different jurisdictions. You might remember that in the '90s, when you traveled around the world, you needed three chips to be able to make calls in Japan or in other parts of the world. So it's not like this is a totally new thing. This is actually a much simpler problem to solve if, at the end of the day, there are two standards we need to manage.
But it'd be nice if we could agree on one, or on some common suite of standards.
Yuval: As we get closer to the end of our conversation today: you mentioned these standards, and to get them in place, I'm sure there needs to be collaboration between governments and private companies like yours. Do you think governments are doing too much, too little, or just right to get to where we need to be?
Michele: So far, I would say they're not doing too much. Too little? They could do more, and maybe they will start to do more. Ultimately, though, you need to be careful what you wish for. They need to help internalize the cyber risk that is just being accepted without due consideration.
Because ultimately, we don't have seat belts in cars because of market forces, right? There were regulatory requirements that said, "Look, we're going to save lives if we require seat belts." And so they were required. The same goes for speed limits and driver's licenses. It wasn't market forces that created those.
So with cybersecurity, there has to be a regulatory push to properly internalize the costs of not properly securing things. But the answer isn't to be prescriptive, right? You don't want governments to come in too soon and start prescribing to people what to do.
There are many reasons why that's a bad idea. But we do need governments to start encouraging accountability in a non-prescriptive way. They do need to start requiring the critical infrastructures we all depend on for our livelihood and our safety to have a proper plan to protect against this emerging threat, right?
So I think governments do have a role to play in properly internalizing that. They also have a slightly more subtle role to play in making sure the global markets in cryptography and cybersecurity function properly. They can't be so nationalistic that your only market is your domestic market, because then there won't be enough money in the system to develop the high-grade products we need.
So we do need an appropriate regime for using each other's products and services, so we can have really high-quality products protecting our digital platforms. Those are two examples of where they're going to need to do more. I'm not saying that in the sense that they've been slackers not doing their job, but now that the world's evolving, that is the next phase, where we need governments around the world to step up.
Yuval: My next-to-last question is: if I'm the CISO of a large regional bank, what should I do today? What should I do tomorrow morning?
Michele: If you haven't already, you really need to make sure that there is somebody in charge, and that this is their number one job, not something they do in their spare time.
This is their number one accountability. This is what their KPIs should be based on, and it's to make sure there's a proper quantum readiness plan at your organization. It starts off with one person, but it probably needs to grow. Again, this needs to be their job, and they need to be set up for success. That means they need not just the CISO's support, but cross-functional support at the organization, to make sure they get access to the risk managers and everyone else they need to talk to, to do their assessment, start formulating their plan, grow the team, and get the investments they need to properly transition the platform.
They'll need to engage people outside of the bank, but it's a balance, right? You can't outsource this kind of accountability. You need somebody on the inside who really understands the business driving things, but you also can't recreate everything that has been created on the outside. So you need to properly engage, but you need to know what questions to ask, right?
And how to interpret the answers, and then know what to do next. So you need a balance: a core team, starting off as one person but quickly growing, driving this transition for your organization and engaging more and more with the vendor ecosystem, so they know what tools and services to procure to effect the necessary migration.
Yuval: So Michele, how can people get in touch with you to learn more about your work?
Michele: Probably the easiest way is just to find me on LinkedIn and send me a note there. Michele Mosca.
Yuval: That's perfect. Well, thank you so much for joining me today.
Michele: It's been a pleasure, Yuval. Take care.
