 Others will follow very soon. And then I'm going to talk a little about quantum information science. And it's more a prep because I see quantum information science having a big impact on CIDR. And I know Frank Donesti, who just gave a talk next door on 5G, some of you might have been in that talk. I have a great admiration for Frank. He's the CTO of the Air Force. And I'm glad to see him in this 5G region in the standard settings that are involved in the testing and experimentation that's going to make major changes in the way we do business. But Frank is going to have a talk on quantum later. He understands this as best as anybody in the world. So this is probably a prep talk to give you the basics before Frank's talk on quantum as well. So with that, let's get going. And what's the right way to look at cyber today? Prep for tomorrow, there's defense in depth. I firmly believe that a good cyber solution requires hardware working with software, resilient hardware. And by resiliency, I'm referring to the NIST standard of 800-193, which is hardware resiliency. And if you're not designing your systems with hardware resiliency in mind, then you're subjected to going down and staying down. So the way I think of it, because of my age, I think of it at the Timex commercial, where the diver from the cliffs of Al Capocco will hold a Timex watch on his arm and dive 200 feet down into the water and come back up and say, hey, it still works. It takes a licking and keeps on ticking. That's resiliency. When you get hit with a cyber attack, because by golly, systems are going to get hit. The adversary is good. They're using AI techniques. And they are spending a lot of money to learn how to get into your systems. So when you get hit, you identify it, you throw them out, and you're back up running within seconds, not minutes, not hours, not days. But if they do get in, they're out immediately. That's resiliency. And that's what the NIST 80193 gives you. Gives you that resiliency. But the defense in depth is good too, so that if one thing is penetrated, you detect it with the next detection device. So that's just general motherhood and apple pie here. We had a research firm kind of look at 350,000 new malware attacks every day. New. Now that's generating 365 days a year. How can they do that? That's just phenomenal. I don't know, but I believe that number when I first started it. And then you realize that they're not really creating absolute new malware. They're tweaking old malware. And they're using AI techniques and using script kitties to do minor tweaks, an old malware, so it doesn't look exactly the same. And by doing that, 99% of them are not gonna work. Because we know what the old malware looks like. We know what it did and we recognize it. And chances are the tweak just wasn't effective. They're looking for the 1%. The 1% that does work. And so the AI capability to go tweak some things gives you speed. Cyber of tomorrow is gonna be all about speed. The speed of the defense has to equal the speed of the offense. And I know it's football season. So it's the same thing in sporting events. If you can't keep up with your opponent, you're gonna fall behind in the score. And sure enough, the AI techniques of manipulating former malware is one of the things that's gonna let that adversary get a step ahead of ya if they can design a tax faster than you can defend against him. So that was the driver that got us thinking about how can we use AI to look for methodologies to defend against future malware that hasn't been created yet? And that's what I think you'll see as the future. All right. What we call it, this is a marketing term. I'm not a marketer, I'm a technologist. So I'm the CTO for HP Federalist Group. That's a portion that deals primarily with the government. I do that because I've been working with the government ever since I was 18 years old and raised my right hand, got sworn into the service. But the deep learning techniques is what we think is an advantage in detecting future malware or sensing this looks like, feels like, smells like, malware we've seen in the past. And in doing so, you can quarantine it, you let the rest of the things go by and see if you can minimize the false positives and false negatives. So a little progression through, those that kind of grew up with compute, my first computing experience was right out of school, I took a course with the GI Bill and how to build a microprocessor, 8088 machine. I learned how to solder, I learned how to put it up, I learned binary code and I could abstract, multiply, divide as good as any calculator with this machine. But I learned the basics of how a computer works. So by the time I got to my first assignment, I was put in charge of a PDP-8, which is a digital equipment mini computer that was collecting sensor data and analyzing sensor data. I don't remember the memory, I think it was 2K. And it ran on some big disc and it had these big toggle switches that you had to line up with the right zeros and one up or down to get the thing booted up. I ran into a lady at the New York Strava conference, a data science conference, who I'd known, probably one of the world's best data scientists, her name's Hillary Mason. She ran fast forward labs and then got bought out by Cloudera. So she's at Cloudera now. And Hillary said, hey, I used to work on PDP-8s. And I said, yes, ma'am, I used to, back when I was a young officer in the service. She said, do you know where I can get one? I said, you know, I'm sure they don't sell them anymore, but there's probably some around somewhere, you know, some basement or backyard. And she said, yeah, my husband collects antique computers and he's missing the toggle switch on his PDP-8. And I thought, yeah, this is really a state of the art here. Some of the things you worked on are now in the museums and the antique stores. So believe me, you'll be in that same position. The second lieutenant's here, give you 30 years and you'll be looking at some antique museum computers that you used to work on. So good luck to you. But in the legacy, the antivirus, Norton, Symantec, the antivirus systems, McAfee, they were looking at signature, signature-based virus detection. So if the zeros and ones lined up in the same order as a same known malware, it was not allowed on your machine and it was prohibited. They evolved to machine learning. And if you think of how AI works, AI is a broad topic. I've been working at AI's for about 40 years. And back in the start days, there were two forms of AI. There were weighted algorithms and there were expert systems. And think of expert systems as kind of like what Watson was when it won Jeopardy. It was able to go through and because of known input, it was able to go and retrieve something that was known. A weighted algorithm, think of it like a Bose headset. You're getting sound waves in, it's analyzing the wave and putting out a reverse sign or a reverse signature to cancel out the sound. And so that's weighted algorithm. That's what AI meant to you. And they were using some vision to put car windshields and the cars and the production line. There was some work in computer vision that was quasi AI. But today, I can probably count 32 different facets of AI all the way from natural language processing to deep learning that I'll go to. But the way there's structured data, there's unstructured data. But in the AI world, you go to machine learning. And machine learning, probably the way to best think of machine learning is try to discern a picture of a dog or a cat. And you put 700 pictures of dogs into the algorithm, process against 700 pictures of cats. And it's looking at the difference between the eyes, the kind of nose it has, they have a tail, what's a tail look like, the measurements of the tail. And the algorithm, when you get the next picture and you don't wanna tell it, the machine, is it a dog or a cat, it should be able to come out but this is more like a dog than a cat, we'll call it a dog. And that's what machine learning techniques do. There were four grad students in MIT that were given a project at Boston Woman's Hospital to go and investigate mammograms or radiography pictures and see what they could find using machine learning techniques. And this is written up in the Second Machine Age, which is a book by Eric Brinjolison who he and Andrew McAfee were two MIT data scientists that looked into how can you use data to enhance the way we're living, kind of a history of data science. They actually hit on the top 10 of the New York Times best sellers list. Probably 50 years since an MIT professor has had a book on the best sellers list. But they wrote about these four kids went out there with no knowledge of radiography or reading rate, were able to take the data and through machine learning techniques being able to discriminate cancer versus non-cancer. Here's 700 pictures of cancer, here's 700 pictures that don't have cancer. Now what's this one look like? And get it accurate to 92% where the pathologists who do this for a living were hitting about 72%. So almost a 20% increase because they weren't trying to use their judgment. They were just trying to say, what did the numbers say? Now when you put the machine learning techniques together with the learned and trained pathologists, the number went even higher, the 95, 96%. So as an example of how a human machine working together can get a pretty accurate read and from the aspect of cyber protection, machine learning techniques really improved our ability to detect malware. But the difference is you really had to have malware you'd seen before because that's what you're training your data set on, known cancer, known non-cancer, known dog, known cat. And you're developing an algorithm on the knowns so if you have an unknown creature it's kind of hard to say, what is it? And so that's where you go into a deep learning technique. It's a subset of machine learning and it's a convolutional neural network if you know what they are. So there's a feedback loop, it's non-linear. So the algorithm, it's not a linear algorithm but what you're doing there, well we'll get to it. Next slide. So what we did, we took over 100 million examples of malware and someone said it's close to a billion. We don't want it. Nobody was counting. We got everything to get our hand on that we knew was intended to damage your system. And we put that into a very large processing computer and worked the algorithm for a long time and came up with about a terabyte sized algorithm that could accurately detect new malware. We downsized that to fit on an endpoint, on a laptop, on a desktop, on a printer, even a 3D printer, which we are making now, and got that down to about 100 megabytes and that's the agent that you can put onto the machines. So the end result was we're hitting about 99% detection of new malware that hadn't been detected before. So the sensor's not perfect. It's not gonna stop everything but it's gonna stop a lot of the new attacks that are coming in on a daily basis. The latency's an issue. If you're gonna do something like that, you gotta worry how this is slowing up your system. If you're working in real time systems, 20 milliseconds may not be good enough. Real times are usually three to four milliseconds. So it is a slow up but you can tailor that down to save more false positives and kind of reduce your latency. And the CPU load, which is a driver indicator, it's about a 1% add to your CPU. So it depends on what kind of CPU you're processing with. These are out on our lead books now. Really working well and you'll see more of that as it comes in the future. All right, so layers of defense. So these are kind of things that we think of in the layers of defense. The threat protection was the sure-sits that just described. Sure-click is another layer. You've heard some tests. In fact, the General Skinner just was talking about how they went to Ucom and did all these fishing examples. And 10% of the people at Ucom clicked on the fish. They went fishing, reeled them in. And that's probably about average. I don't think that's unusual. Nothing against the people in Ucom. You can make these things look really real. I know I've got one from Bank of America. I would have sworn it was an email from Bank of America. It wasn't. And the only thing that caught my eye was the address of the at BOA because it had some extra letters in there where they had screwed up the email address for the return address. I didn't click, send it in. Sure enough, malware. But it looked and feel, they'd gotten the graphics and made it look like a Bank of America email. And since I banked with Bank of America, my instincts were just to click. What the sure-click does is when you click on our elite book machines, it brings it up in a virtual machine so that it will operate in its own environment. And if it's gonna go off-site and download malware, that malware is then downloaded to the virtual machine. And when you exit out of your browser, it disappears. So it goes away. Windows Defender. Any of you use Windows Defender? Some of you do. Some people use McAfee or other things. They're all good. But Windows Defender has gotten better. And I'm impressed with the way they're operating right now. And a lot of people over the years have made fun of Bill Gates and Windows. The hackers all wanna attack the Windows because there are a whole lot more of them than there were of OS systems. Up until the phones came out, that was true. But the Windows Defender, because of the investment Microsoft has been making, has really improved the protection that I give you. And that's what I use. I don't add on the McAfee per se. You can. I'm not sure it provides any extra benefit because we invest about a billion and a half a year to improve the cyber that are involved in our products. That's a significant investment. And we focus just in our products. We're on a sixth generation out in the field today. The seventh generation's in test and we're planning a eighth generation. So every year we're trying to roll out improvements. So the sure sense that I talked to you about on your deep learning, your little networks, it will be better next year and even better the year after. Plus we're gonna invest to make this better and find out what we don't know and learn what got through that we should have stopped and get better. And then our end point controller, this is probably the thing that I'm most excited about is the BIOS protection from using a hardware software combination where we have an encrypted hardware BIOS that is cryptographically signed and put in to each individual machine. And it checks what has been loaded on boot into memory. And if there is difference during operation, it flushes memory and it reloads it. And that's the definition of resiliency. That it can take a hit, you can get a low jack attack or something that's coming in under the operating system. It will be detected if it's attacking the BIOS and it will be removed and replaced. And that's the kind of speed of defense that's gonna keep up with the speed it offered. So you might get 15 to 30 seconds in the machine to do something, but it's not gonna reside for 189 days as a persistent threat underneath the operating system. That's what we were trying to stop there. That's a little bit of sense that just says, hey, this stuff can work, this is the state of the art, but it works offline. So quantum information science, this is where it gets exciting. There's probably $20 billion a year invested in quantum research around the world. And this is the message I really wanna get out is that 20 billion a year, only about 477 million is from the US government. That's a very small sum. Now there are other classified programs that Dana Deasy has discussed with me that they are investing in in the services. He's a CIO for DIOD. He understands this problem and investment. He would like to have more money, but you still gotta buy planes and tanks and ships. And so there's a constraint to any need that the services have. There's a rest of government needs to invest in this National Science Foundation is doing a pretty good job with a quantum leap program. DARPA has some quantum programs. They've looked at quantum annealing and they're looking at some other things. But the quantity of funds, Australia spends about twice what we do. The UK is about equal to what we do. The European Union is about three times total in quantum information science. And I'm not talking about a quantum computer. Quantum computing is a whole nother field. It's a subset of quantum information science. But there are things of knowledge and quantum, we'll cover them, that can be scary. The Chinese invest 10 billion a year in quantum information science. Most of this in quantum communication. Some of it in quantum key distribution. We'll talk a little bit about this as well. But when you realize that a Chinese researcher cost 80% less than a US researcher doing the same thing, they just don't pay them that much. And yet, you know, they're happy to work there. They don't have the competition we have in the US by industry trying to grab them out of the lab, out of academia, and put them to work in product development. And because of that, that 80% less means that the 10 billion US that they're spending is equivalent to us spending 50 billion. And yet, we're spending 477 million. Three orders of magnitude. Really, almost three orders of magnitude less. So what should we be doing? I mean, all we do is read Chinese literature. I don't read Mandarin, but I kind of wish I did. That would be something. And look at when the science heads from science to time to productize some of this thing. And if we can be the first to patent and product, then I think we'll be okay. So that's kind of, what is it? What is quantum information science? Is it ability to use macroscopics or microscopic scale, atomic information to carry information? All right, so this is something that I worked on the last four or five years with a company called Cubatec out of Carlsbad, California. There are several companies that do things like this. And it is a quantum key distribution capability, but they make a kit and they sell it to high school physics lab of how you do entanglement of photons. And so by entangling photons and sending it to two different ends of a fiber optic wire, you can show that you're gonna get the same results every time. That the two photons are gonna have the parallel created in the same manner. And what they're looking for primarily, the kit says $50,000 for your high school science fair. That was years ago. I think they're down to about $10,000. So you're making enough of these and putting them out and the high schools wanna make sure that their students are up to speed and some believe it's science that these things are coming out. And they're looking at polarity. So a light ray has a polarized face and that polarized face is looking at either a vertical or horizontal polarization. It's like your sunglasses. You wear sunglasses. You know, you're either vertically or horizontally polarized. So only half the light is through. Well, that's kind of what they're looking through here. All right, so classic states that quantum can use is the spin of electron charge plus or minus polarization and there's several other factors. Electron spin is either up or down in an atom. And so by just measuring a spin, if it's not up, you count it as a one. If it's a down, you count it as a zero. These are some of the basis of using quantum information to carry digital information. This is a block sphere. So we're gonna start to get to what a qubit is. A block sphere, instead of having a classic bit of being a zero or a one and that can put in enough zeros or one in a string to carry in information, get you a binary code. A block sphere is, it can be anything from zero to one or anything in between. So you start looking at vectors. When you start talking about quantum computing, you're looking at things like a block sphere. So you can simultaneously calculate every position and calculate that in, it's called a superimposed time or in faster than real time. And that gives you the speed to do things like break encryption. So SHA-256 encryption. I think it was designed to be a hundred years brute force unhackable. I did some back of the envelope computation and said, if you really had use Schor's algorithm, which is a quantum algorithm with a quantum computer, that could probably go down to four hours. So if you're encrypting things at four hours later, somebody can read. Is it really encrypted? I've been told by some real experts that four hours is really not the right calculation. You miss something in your numbers. It's more like four seconds. But it depends on your assumptions. And so you're wasting your time to encrypt it if it can be decrypted by a machine four seconds later. Fortunately, we don't have a quantum computer today. Within five to 10 years, we will. And then certain countries are collecting classified encrypted data that we produce because they know in four or five years they'll be able to read it all. So that's the risk. Now the good news is NIST is working to replace this. NIST is working to make a quantum resistant encryption capability to replace SHA-256 or SHA-512. And they started with 70 algorithms. They've necked that down to 26. By the end of this year, they hope to have five remaining candidates. And by next year, select one or two. That would be the finalists. So they're testing to say, hey, if we had a quantum computer, if we had something that could do this kind of processing, how good would this new algorithm be? And there's some techniques out there that are pretty impressive. OK. There's a problem in measurement in quantum systems. And once you measure it, it changes the state. So superposition is a concept that you can be in two places at once. That's part of your entanglement, two things that are entangled with each other, the same thing, but in two different places. But as soon as you look and measure, then it's fixed. And then any of you take physics and study Schrodinger's cat. Schrodinger's cat, Heisenberg, came up with this concept. Anybody watch Breaking Bad? Heisenberg? So that should be a familiar name, New Breaking Bad. So Schrodinger's cat, Heisenberg came up with the principle that the cat is in a box and it's either alive or it's dead. And you don't know which. But it's going to be in state one or state two. The electron spin is going to be up or it's going to be down. And you don't know which until you look and measure. So you open the box and then you see. But once you see, that's the state it's in. You can't change the state. Well, I guess for the cat, if it was alive, you could always kill it, but Schrodinger Heisenberg didn't build through that. Anyway, this just tells you how you measure and think of it like a photon, a particle, a light that is being measured for polarity. It's going to be a 0 or a 1, something in between. It's the in-between cases you have to learn how to deal with. And one of the interesting things in physics is at the quantum level, is the light a particle or is it a wave? If you can create a single photon, is it a particle or is it still half the properties of a wave of light with multiple particles? And the answers the physicists will give you is both. Now, that seems like a cop-out to me. But I tend to think of it more as a particle effect than a wave effect. But it's still vibrating like a wave. So it still has vibrational energy with polarity involved. So this is a study on entanglement that if I impact a electron here 12,000 miles away, if it's entangled with an electron there, that they instantaneously will change state at the same time. That's a mind bender. I would tell you I understand it, but I don't. And even some of our best physicists of day say, it's a concept, it's a theoretical principle, it has been proven in experiments, but it's really not understood. But the thing that really trips people up in this entanglement issue, if you could really do that, the instantaneous change means you travel faster than the speed of light. The change is occurring faster than the speed of light. And that's what gives rise to some of these science fiction novels of how you travel faster than the speed of light. So not really relevant to what we're going to deal with in cyber today, but it's an interesting fact that might be important. We've already talked a little bit about qubits, and I won't go in any more of that. But it's really, there's a real component, an imaginary component. I know any electrical electricians out there, you know the three-finger rule, current motion flux. And so that's the way I kind of think of the qubit. So here are six fields in quantum computing. I'm going to add a couple of more that are already here that are not on this list. Quantum key distribution, that's where you can entangle two photons. You send it down 20k fiber optics. It's red on polarity. Horizontal is a zero, vertical is a one. And you create a key, not a 256-bit key. You create a key as long as the number of photons you send out. So it can be a million-bit key. There's no limit. You just collect the zeros and ones until you want to say, after a million, that's my key. And you don't need a public key. So there's no public key for structure that you can break in. And the man in the middle attack kind of goes away. The only way to hack into this system is to get physical access to the fiber. And when you do that, when you get physical access to the fiber and you inspect to see where that photon is a zero or a one or a horizontal or a vertical, then it interrupts and changes the polarity and you don't get the same thing at the other end if you get the light at all. It's a single particle. The particle goes to the hacker. And so the key won't work between the two sites. And so it's a pretty foolproof key. Now, if you're at one of the endpoints and see the reading at the endpoint, you can physically be there, then you can determine what the key is. Well, what the heck? If you're at the endpoint, you're not gonna be reading the key. You're just gonna be physically breaking up the gear. And so it's been bought by the San Diego utilities to protect their SCADA devices. And it has recently been sold to Pacific Gas and Electric to cover all California utilities. Of course, Pacific Gas and Electric are bankrupt, so I'm not sure they're gonna pay for it, but they did see the value in SCADA protection on a facility. And they're gonna distribute it, I think, within the next two to three years if they can scale, Cube Attack's gonna be having these systems all over the world. We'll see. Quantum materials, there's some use of quantum science in material development to get properties you want. This is particularly being used in the pharmaceutical industry, so not necessarily tied to cyber. Quantum sensing's being used in systems today. Anybody heard of a squid? Super cool quantum interface device, a very, very sensitive detector of magnetic fields. So you can detect to the Tesla level of a magnetic field. And that is being used potentially could be used as a navigational replacement for GPS. I don't know that they're headed that direction or they're using it, but if you could know specific magnetic field strength in any location in the world, then you can quickly try and be like that as you move. So the simulation that quantum can do, quantum simulation is be very useful in process development or chemical reactions or things like nuclear blasts and simulations. And the quantum compute we talked about the two that I didn't mention were quantum memory. If you can store things at a quantum level, you could put the Library of Congress on the head of a pen. Once again, my analogy breaks down when the experts tell me, no, Gardner, you're all screwed up. You can store hundreds of Library of Congresses on the head of the pen at the bottom. Yeah, and that's because of Avogadro's number of atoms. If you really could create where atoms created this digits for memory, you could do that. And the quantum random number generator from an encryption point of view, this is probably the most important. If you get a true random number generator, which we don't have today, then you've got the ability to get fully homomorphic encryption. And if you could create this device that's in test right now from a company out of Virginia that creates a true random number generator, it's gonna change our encryption processes. NIST is aware of this, they're looking at this and I think it's got great potential. All right, so this is, I think I've described this how quantum key distribution works, fiber optics, you got a sensor on each end, you got a transmitter that's entangling two photons and send them in two different directions. Slave and a master. The Chinese have claimed to do quantum key distribution over a length of 1,200 miles. They claim to have faster than light communication. The truth is that the entangled pair they're trying to distribute over that distance, it works about one out of a million times and it only works at night. So the Chinese are successful in the test, it doesn't work every time. So the functionality's not there, but if you're gonna spend equivalent 50 billion a year, chances are they're gonna start to work through some of these problems. So it's a risk if you can get to quantum faster than light communication and there's a question of whether it's truly faster than light or not. This is the quantum key distribution, so very much like what Cubitech does, but what they're doing right now that Cubitech is not is amplifying the signal. So to amplify a particle of light, so it can go further than about 20K and make it go 1,200 miles, that's a trick. And there's techniques that may be able to do that, we don't understand them yet, we're learning about it, but to do that without losing the polarity they originally had, that's the trick. And there are people working on that right now. Quantum sensing is, can you go down to that atomic level and sense, I think this is something that's most likely to be involved in compute long before a quantum computer is. You'll see a hybrid computer if you think of the days back to my example of the 8088 chip, and if those of you remember having a computer like that, you had the option to buy 8087 math coprocessor and the math co-sprocessor accelerated the calculations enough and so in the quantum hybrid you'll have a conventional workstation that'll have four to eight qubits that are associated with the workstation and you'll offload the heavy math to the qubits and so it would be like the 8087 math co-processor. And quantum sensing, here's the Lockheed satellite less navigation, it's based on a diamond flake nitrogen vacuum sensor, so it's very similar to a squid. All right, so here's the revolution you're gonna see in quantum information security, that's the big deal, that's what impacts cyber and the future of cyber sensing and measurement, materials are gonna change, new materials are gonna be there, room temperature superconductivity is one of the areas we're looking at, if you can get to that and you can start to build a quantum computing at an economical cost because you're not pouring millions of dollars in just to keep the thing super cool every year. And I think the quantum random number generators can be the biggest impact of all and that's probably within a year you're gonna see a commercial product in that field. So if you're interested in learning more, I recommend to you Dr. Phillip Ball, he's a rights for nature but he's got a YouTube video beyond weird, pretty easy to remember, it's written down here, everything you thought you knew about quantum physics is different. He's entertaining, he's understandable and he gets great examples and I've tried not to steal any of his examples but I encourage you to listen to his video, he has a book, book's a little longer to get through in a more detail. In the video he doesn't use a single equation but he helps you understand the concepts. So I'd refer to that, there's some other things here if you really wanna get into this and other than that I'll stop and we've got about five minutes for questions if anybody wants to questions or first question, no. I've got a coin, if you wanna challenge a coin, that's fine, here's a question, yes sir. Cubits or something like that. Yeah, IBM's working on it, Google's working quantum, they're using ion trap methodologies, there's seven different cubic potential forms or factors that NSF is tracking. D-Wave is using quantum annealing and IBM's claiming they've got about 2,500, 2,600 cubits. The truth is with the ion trap methodology, they only last for like a quarter or three quarters of a second and then it disappears. So you gotta do all your calculations in a very small amount of time and the 10 cubits are required to error correct one that you're using to calculate and so because there's so much noise in the quantum fields, most of the cubits are used for noise reduction, error correction, or getting the signal noise ratio right. So there is some investment and that investment is not included in the 477 million that NSF says the US is investing. So we do have private investment that's adding to the, you know, the 20 billion over the course of the year. So that's good. Let's wrap it up there, I'm down here for a while. Come get your coin and thanks for your time.