Root Causes 65: Quantum Key Distribution

Tim Callan

Today, we are here to talk about quantum key distribution.

Jason Soroko

Yeah. It comes up. It’s an interesting topic. I think the first thing, Tim, we have to say is we’ve done a lot of podcasts on the topic of quantum resistant algorithm, which are essentially alternatives to the algorithms that we know and love, RSA, ECC and those kind of things and quantum key distribution is not that.

Tim Callan

Right.

Jason Soroko

It is kind of like quantum random number generation. It’s a whole thing unto itself.

Tim Callan

I was just going to say, we keep getting questions about quantum random number generation and it makes sense because there are all kinds of crypto stuff, and they all have the word quantum in them. Same thing happens here, right? This topic keeps coming up and even though they are not connected maybe this is a good time to clarify what they all are so that people understand this landscape of words that start with quantum.

Jason Soroko

Yeah. Quite often, the word quantum especially with respect to quantum random number generation or quantum key distribution, the reason why the word quantum is used is because it’s using some physical state that is unique to quantum - - basically the properties of a quantum mechanic system. So, in other words, quite often you are dealing with the special properties of light or something like that. When you are talking about quantum resistant algorithms, you are really talking about new forms of math meant to be able to be resistant to quantum computers, which in themselves are using some form of quantum mechanics properties as well.

Tim Callan

Yeah. So, there are three things in this case that are all operating differently based on the facts – the essential randomness of quantum mechanics, right? And it is quantum computers. It is quantum RNGs and it is now quantum key distribution and that’s where they are all connected and that’s where the word quantum comes into that but then of course because it’s key distribution and random number generation people naturally attach it to these kind of PKI topics.

So, all right. Quantum key distribution. At a high level, what is it?

Jason Soroko

It’s key distribution.

Tim Callan

But it’s quantum.

Jason Soroko

But it’s quantum. Absolutely. So, it might actually be more interesting, Tim, to talk about what we are trying to solve and then we are going to get into an answer to your question of what it is.

Tim Callan

Great. So, let’s do that. So, what are we trying to solve?

Jason Soroko

Yeah. So, if you think about side channel attacks against traditional methods of key distribution. So, RSA 2048 still considered safe but there have been some people, especially with lesser bit rates, etc. actually trying to attack it with side channel attacks, which is in other words, listening to, goodness, sometimes it could be audio. Sometimes it could be just electrical generation of when people are trying to solve a math problem associated with the key distribution. All these kinds of things can sometimes leave a certain artifact which can be picked up by some side channel attack. And those are real concerns. In fact, we have probably read a lot of these things recently about side channel attacks inside of Intel chips and things like that. Those have been interesting attacks recently.

Tim Callan

Yeah. So, kind of to put in layman’s terms, you would measure something that is measurable about the hardware that is doing the computing. So, like for instance, the power consumption. And based on how the power consumption changes, you can draw certain conclusions about the nature of the work that’s being done and in so doing you can, what do I want to say, reduce the potential key space that you might actually be dealing with and it makes it conceivable to then reduce the key space down to the point where you could actually do a brute force attack.

Jason Soroko

Right. So, and in this case of quantum key distribution, we also have to further say that we are really not talking about the encryption here. We are talking about the generation of a key ahead of time between an Alice and a Bob. You know, how do you begin to create a secret between the two of you so that then you can go further in your communication and actually encrypt a message between the two of you.

Tim Callan

Yeah, and in one of our recent podcasts you mentioned an idea to me, which was in IoT devices using the inherent flaws that exist in the silicon as a source of randomness to make a key. So, the idea here is to say, well actually, all matter is a source of randomness. Right? So why don’t we do that instead.

Jason Soroko

Yeah. And in the case of quantum key distribution, I guess one of the things that’s being attempted to be done here is to actually utilize properties of quantum electrodynamics essentially to be able to establish a secret between each other that is extremely difficult to either have a man-in-the-middle attack, a side-channel attack, you know, if somebody is eavesdropping in even if they were able to take measurements of the communication of the quantum key distribution how difficult can we make it in a quantum state. And that’s what quantum key distribution really is all about.

Tim Callan

All right. So, without getting too deep into physics lesson, how does it work?

Jason Soroko

Right. Well, there’s two kinds of quantum key distribution. The one I think we will be talking about the most here today is the prepare and measure method of quantum key distribution, which takes advantage of the fact that quantum indeterminacy.

Tim Callan

Ok.

Jason Soroko

And, of course, you might have heard of the Heisenberg uncertainty principle. Right? The big argument between Einstein and Heisenberg, etc., and keep in mind, the most important thing to keep in mind about this is that these various states, these various quantum states cannot be measured without disturbing the original state.

Tim Callan

Right. Correct.

Jason Soroko

And if you can take advantage of that, that means that if you can send say photons that have been specially prepared and you have a way for the legitimate receiving party to then take measurements of that in such a way as heck, you know, maybe there is an interloper who is also listening, that quantum indeterminacy actually is interesting because you can then detect whether or not those photons have actually been either looked at by someone else or have been messed with in some other way. And so, therefore, it’s an interesting way of using quantum electrodynamics in order to be able to share a secret.

Tim Callan

They stop being waves and they start being particles and you can say, hey, wait a minute, my waves are particles. Who has been looking at them?

Jason Soroko

Not quite, but you get the idea. And the other way, just to be completest about the notion here, is there is also entanglement-based quantum key distribution methods as well, which that’s a whole other subject that has to do with - - yeah, that’s another - -

Tim Callan

That’s even further out I would think, right?

Jason Soroko

Yeah it is.

Tim Callan

Entanglement of course is a super exciting idea because among things it appears to get around the problem of the speed of light but it’s - - I mean we are a long way from practical entanglement based systems. Like a long, long way.

Jason Soroko

That’s right. And we are dealing with a different property here. This is not just quantum indeterminacy, this is also dealing with the fact, the amazing fact of you can - - in simple terms, spin a particle in such a way that, you know, to spin two particles, say a split photon in such a way where regardless of how far those two split halves get away from each other, if one half starts spinning in another direction because of some - -

Tim Callan

Yep, and then you read it on the other side. But, that’s quantum entanglement. That’s not what we are talking about.

Jason Soroko

That's right.

Tim Callan

And so, I don’t want to say, this is still very advanced science but a simpler quantum key distribution system. So basically, I fire a bunch of photons down a fiberoptic cable and the party on the other end is reading those photons and because of the fundamental quantum randomness, they’re going to get an original unique random read at that time that will never be the same as another time that you fire photons down a fiberoptic cable.

Jason Soroko

Close.

Tim Callan

Basically correct?

Jason Soroko

Close. Close. Let’s talk about a particular protocol. It’s actually fairly well-known. Bennett and Brassard. Their 1984 protocol. I think it’s called often BB84. Right.

Tim Callan

Ok.

Jason Soroko

And exactly, right, Alice, would be sending specially-prepared photons down fiber optic cable and Bob will receive them. But, what’s interesting about this is two things. So, for each photon that’s sent, that would represent a bit. Right. Server 1.

Tim Callan

Right.

Jason Soroko

And Alice will randomly select a basis. Now this basis is a pair of orthogonal states and the two choices that Alice could make is either it can either be rectilinear or a diagonal state. Right?

Tim Callan

Ok.

Jason Soroko

Which is essentially two orthogonal halves of each other. And that’s a very key point to this and then a polarization state is chosen. This has one of four possibilities, which is of course, zero degrees, 90-degrees or 45-degrees, 135-degree. So, it’s in one of those. You can imagine that in your mind.

Tim Callan

Yep.

Jason Soroko

Now Bob at the other end, for every photon that he is gonna measure will actually randomly guess at the state of each photon. In other words, Bob will actually randomly guess at the basis – that orthogonal - -

Tim Callan

So, he says, I think this one is 45-degrees.

Jason Soroko

Right.

Tim Callan

And then he sees if he is right or wrong.

Jason Soroko

That is exactly right.

Tim Callan

Ok.

Jason Soroko

And so, then what happens is Bob then sends all the measurements back to Alice and then Alice sends back all of the correctly chosen state basis.

Tim Callan

Ok. And then they use those to make the key?

Jason Soroko

That’s it. And the ones that don’t match are just discarded and that’s usually about 50% on average.

Tim Callan

Ok. Sure.

Jason Soroko

And then the final part of this is kind of interesting in that in order to check for eavesdropping, Alice and Bob actually compare the predetermined set of remaining bits.

Tim Callan

Ahhh.

Jason Soroko

And if there’s a certain amount of error in the assumption then what happens is they know, oh geez, I think Eve must have listened in on us and was taking measurements at the same time.

Tim Callan

And then we throw it out and we do a new one?

Jason Soroko

Exactly right.

Tim Callan

So, some questions in this. When we talk about this. So, at a high level, to the degree that anybody can understand matters of quantum physics, this makes sense. But then it invites a bunch of questions. Like, you know, if I’m firing a bunch of light down a fiberoptic cable does that mean I have to have a direct connection to you? Does this work like in a traditional kind of internet environment where stuff is bouncing across multiple servers or is this built for a specific connectivity scenario?

Jason Soroko

It has to be very specific because we are talking about photon. So, if you think about it, you know, this isn’t just data. It’s actually the physical quantum electrodynamic property of a photon, which enables you to have this incredible physical property of those orthogonal states. I mean we can definitely get into the really interesting math of this, but it’s the physical photon. It’s the thing itself.

Tim Callan

Right. So, I have two physical locations that I want to communicate with each other and I have ultra-high security and I run designated fiberoptic cable between the two and that cable is there for purposes - - it might be for other purposes, too, but at least in part for purposes of allowing me to generate these keys and literally it’s going from Site A to Site B and I’ve planned ahead of time that Site A and Site B are the two places - - is that correct?

Jason Soroko

Yes.

Tim Callan

So, this is a very - - even if all the engineering was worked out, this would be a very, very corner case kind of application?

Jason Soroko

It’s so corner case - - I think it’s Bruce Schneier wrote a really good blog about this and he was quoting GCHQ from UK. And as he said, they delivered a brutally blunt assessment of quantum key distribution.

Tim Callan

Ok.

Jason Soroko

Yeah. And of course, you know, quantum key protocols - - quantum key distribution protocols really only are addressing the problem of agreeing for key types of encrypting data. Everything else - - like we’re not even talking about anything else. We really are just talking about quantum key distribution. A key distribution in general. And so, therefore, this is absolutely not replacing flexible authentication mechanisms that we know and love. In other words, this is not going to work for IoT.

Tim Callan

Oh goodness, no.

Jason Soroko

It’s not gonna work for Big Data. It’s not gonna work for social media. It’s not gonna work for cloud applications.

Tim Callan

It’s not gonna work for commercial applications. It’s not gonna work for your average enterprise. This would have to be like military or government or the most secure of secure before you would be doing this.

Jason Soroko

It’s incredibly clever. Right? This is Bruce Schneier. I’m paraphrasing. It’s a clever idea but basically useless in practice. It’s so niche that you almost have to - - probably you and I are not privy to where this might even be used. In fact, if you look around there’s companies out there like ID Quantique out of Switzerland and there are others as well. I could name them, but you can look them up and I think the financial industry, you can imagine, they might want to send large amounts of money between each other. Or, something that’s secret. And definitely government and military, but beyond that, we are just not gonna see this.

Tim Callan

Yeah. Yeah. Exactly. You start to say, look, I’ve got different knocks that are located not located too far, a couple blocks away from each other in Manhattan and they are connecting with each other and if someone hypothetically could get in there and do what they want to do they could siphon off billions of dollars and there’s kind of no amount of money is too much from a security perspective is the posture and gee, maybe we could lay this cable and maybe this would prevent some kind of real sophisticated ATP style attack. Right? I get that. Or APT style attack. Excuse me. I get that. And, maybe that’s the only kind of application that there will ever be for this. I also wonder, like running long fiberoptics, like how far is practical? Am I using this to go across the street? Surely, I’m not using this to go across the country?

Jason Soroko

I think you - - you know what? This is where my knowledge of this goes a little hazy but my assumption is with enough amplification potentially you could.

Tim Callan

Ok.

Jason Soroko

If the fiber is there, it can be done. I’m sure somebody who knows more about this could tell me the exact, you know, limitations of this, but if the fiber is there it can be done.

Tim Callan

And then of course you are defending the, depending on the physical aspect of this fiber, right? If anyone with a backhoe can take you down then that’s another problem that needs to be overcome. So, you would need to make sure that this cable was in a physically secure environment for its entire length.

Jason Soroko

And here is this other problem, Tim. This might be the more interesting problem to the listeners of this podcast, which is if you are Eve, you know what your limitations are. Keep in mind that your job as eavesdropper is to be able to make a copy for yourself and also send a perfect copy to Bob.

Tim Callan

Right.

Jason Soroko

As to not trip off the warning.

Tim Callan

Right. Cause otherwise Bob knows and he doesn’t use it and you’ve got nothing. Yes.

Jason Soroko

Yeah.

Tim Callan

You need Bob to use the key.

Jason Soroko

Well, without getting too much into the math, Tim, what I find interesting is there is entire endeavors of study of how Eve can do approximation and so therefore, even the way that Alice has to make choices about which orthogonal states that she is choosing have to be done very, very carefully otherwise, Eve could make very good guesses as to what is being said.

Tim Callan

Ahh, got it. And then if Eve is making a very good guess, even if it’s not right every time you could say, well what’s the acceptable level of it could be right and if Eve’s out there guessing at your certs and your keys and maybe you are broken and you are never gonna know that you are broken then you can’t use this.

Jason Soroko

Yep. So, in other words, nobody here is saying that quantum key distribution algorithms are inherent perfect. Right? Nobody out there is saying that at this point.

Tim Callan

Ok. All right. So, this is definitely gee-whiz advanced science. It’s good to know about. It’s always fun to learn about stuff. And, like I said, it’s a topic that comes up and you understand why, but it’s good to clarify for the listeners that this not really part and parcel of our sort of what, what do I want to say, our every day production, PKI systems that are running the digital world in a way that for instance, quantum safe algorithms are.

Jason Soroko

That’s absolutely right. I think that eventually one day we will probably be running some form of quantum resistant algorithm. That’s probably a reality and that will be in response to the fact that quantum computers will get a certain state.

Tim Callan

Yep.

Jason Soroko

In fact, it’s almost every few weeks now where we hear some new form of quantum supremacy where Google or IBM or Microsoft or one of those guys, you know, is declaring themselves as being top of the heat and I think we are gonna continue to see that for the time being.

Tim Callan

Right. But meanwhile, quantum key distribution not really playing a role in that.

Jason Soroko

No. But there are commercial interests in this because of the fact that there are some levels of secrets out there where this particular niche technology is not just interesting, but perhaps useful. So, for those of you working in the field – fantastic. But for the rest of us, I just don’t think we are gonna see it.

Tim Callan

Yeah. Yeah. So, it would be a very small percentage of the use cases and applications and the individuals who work on this stuff who are actually gonna be touching that.

All right. Great. That is a great cogent explanation of quantum key distribution. You know, I had a very light understanding of the topic until we came into this and now, I think I know what I need to know. So, thank you, Jay.

Jason Soroko

Yeah. Thank you, Tim.

Tim Callan

All right. And thanks everybody. This has been Root Causes.