 webinar series of the IT Journal on Future and Evolving Technologies. My name is Alessia Magliarditti from ITU, the International Telecommunication Union. Alessia, we cannot see you. Sorry, sorry. I repeat again, hello and welcome to the webinar series of the IT Journal on Future and Evolving Technologies. My name is Alessia Magliarditti from ITU, the International Telecommunication Union. ITU is the United Nations Specialized Agency for Information and Communication Technologies. ITU allocates frequencies to the EU, to the services that make use of the radio communication spectrum. It develops standards and assists developing countries in setting up their information and communication infrastructure. ITU and academia share a commitment to the public interest and this commitment is embodied by the IT Journal which offers complete coverage of communication and working paradigms free of charge for readers and authors. Our journal welcomes submissions at any time on any topic within its scope and we believe that this webinar series launched in March this year will inspire more contributions from researchers around the world. It is my pleasure to open today the webinar with Professor Edward Knightley from Rice University USA and we count on your support to make this webinar an interesting experience so please submit your questions via the Q&A channel. We will address your questions to the speaker during the Q&A session and after the talk and the Q&A please stay online for the wisdom corner, live life lessons. Professor Knightley will guide young scholars in the field of current ICT research. Now I'm very pleased to introduce the moderator of this webinar Professor Iana Kildiz editor-in-chief of the IT Journal and founder and president of TRUBA from the United States. Professor Kildiz is Ken Byer chair professor in telecommunication emeritus at the Georgia Institute of Technologies. In the last decades he established many resource center worldwide including in Finland, in Spain, in Africa, and Saudi Arabia. He's editor-in-chief emeritus of impact factor journals and at the top of the most prestigious international rankings. He's visiting distinguished professors in many universities around the world. He's also author of many patents including the last one received on Qt satellites so congratulations Ian and his current research interests include 6G, 7G wireless communication systems, hologram communication, terahertz, molecular communication, intelligent surfaces, nano networks and many other subjects. So Ian the floor is yours for your opening remarks and to introduce our speaker. Thank you. Thank you Alessia and I hope I'm on and I would like to welcome you again to our ITU journal for future and evolving technologies webinar series. Good morning, good afternoon, and good evening all around the world. I have a great pleasure to introduce you. One of the greatest researchers of his generation Dr. Edmund Knightley is our distinguished speaker today. Before I present you Edward's career I want to share my personal experience with him. I met him almost 30 years ago when he was still a PhD student with my dear friend Professor Domenico Ferrari at the University of Berkeley. I personally liked him and also I was very impressed by his research. That time it's also a good test for my memory. I'm sure he will like it. He was working on scheduling algorithms for wired networks. He published many firing research papers that time and many many more afterwards on very interesting hot topics in communications and networking field. Edward received his master's and PhD degrees from UC Berkeley in 1992 and 1996 respectively. He has BS degrees from Auburn University in Alabama in 1991. Edward joined Ryan's University Department of Electrical and Computer Engineering in 1996. Edward went through all ranks and served the Department of EC as the chair for five years and also he became a chair professor at Rice. I have to also say that's my promotional second. I take enormous pride that I supported Edward through his entire career. We had many many personal interactions at conferences and I visited him many times. He came to Georgia Tech many times so it was always fun to be with Edward. His research revolves around networking, mobile wireless systems and also lately security. His research focuses more on protocol design, performs modeling and evaluation and also he has some experimental test bets like the one that I will mention about urban scale like wireless mesh network. Edward and his group were the first to create a multi-user beamforming wireless LAN system and they demonstrated a multi-user MIMO in the wireless networking standard in IEEE A2.11ac. Edward is an excellent speaker and does he's well-sought and invited keynote speaker for many leading conferences. Even I invited him in many conferences as keynote speaker he always delivers an excellent keynote. Edward received many awards. They're not easy to list all of them and some of them are for example, he is a Sloan Fellow back in 2001 and I Tripoli Fellow 2009 and ACM Fellow in 2017. He chaired many many conferences really, leading conferences not like all you can eat conferences. They're really prestigious conferences and he did an excellent job. In 2007 I visited him at Rice University. That's why I remember it very well because he showed me his wireless mesh networks test bet and this was an excellent community service because he was providing internet service to poor communities in Houston. So that was really impressive. That's why I still remember it and also in 2016 a video of nightly's work was featured during the White House's announcement of a new 400 million wireless initiative and the lately the last four, five, six years he had some pioneering results for telehealth communication networks and his Google scholar H index is 98 and total number of citations is 46,632 and I'm sure it's increasing rapidly and again we thank you Edward for accepting our invitation and we look forward to your presentation. Thank you Ian so much. It was for that introduction but also to get to relive some of those wonderful times we had previously. It really I do appreciate so much it goes back to student days too and that's back in the 90s. I don't know if all the years were in the bio but it's been a pleasure to know you since the 90s and also thank you for your support through the 90s and all those invitations for keynotes and so on. Just really impactful so thank you and also thank you for inviting me today so maybe I'll see if I can share my screen so and also I'd like to welcome our online audience and I look forward to Q&A afterwards and you can put them in the chat or audio I'm not sure how we'll do it at the end but I'll be happy to take your questions but so today I'd like to talk about some of our recent work starting with millimeter wave in 60 gigahertz maybe 10 years ago or so and moving up to terahertz in more recent years and so I'd like to talk about the three aspects of that wireless networking sensing and security and targeting higher frequencies so whether 60 gigahertz or 100 all I think my title is 100 but I'll actually start with today's standard at 60 so I changed my title slightly for 60 but I'd like to begin with what are some of the things that we're trying to do in next generation wireless and we have data rate challenges of how do we get to terabits per second while having mobility low latency robust secure from eavesdroppers and attackers but also the fusion of sensing and communications and I'd like to show you the the points to the picture on the lower left as you can tell us pre-covid because everybody was sitting so close together but that's a that's a VR conference and I want you to notice that everyone has a wire coming from their headset and so if we think about trying to make wireless work at those sorts of not just data rates but densities gigabits per second per square meters that that we're not yet there today and so we have farther to go with all of these metrics to support the new applications and whether it's it's a VR AR or UAV network or something we haven't foreseen yet the new capabilities that we introduce for networking security and and sensing can help foster development of those new applications so I'd like to turn to the issue of the the issue of spectrum and what are some of the why should we be looking up at higher frequencies in millimeter waves and and subterrors and and what are the trade-offs that we have and so just to remind you of the freeze equation shown there on the right we can see that the received power has in the denominator this frequency squared term and so the other terms are the the transmit gains of the antennas and speed of light and the distance but I just want to focus on the frequency for a moment and and just ask ourselves well what how much do we lose what's that d-squared doing to us at at 10 meters just to get it in perspective and so earlier in my career I worked in UHF TV bands and those were wonderful for Wi-Fi we called it super Wi-Fi at the time and and so that gives us a baseline of losing 40 60 v at 10 meters and then as we go up in order of magnitude and spectrum to today's Wi-Fi around 5 gigahertz then then we're we're losing in not just the factor of 10 but the factor of 10 squared and so that's where you know these these 20 db differences come and then we go up to millimeter wave and now we've lost about another 20 db and beyond so these these you know losing a few db here and there might not be such a big problem but in the end we've we've lost quite a bit of gain as we as we go up to higher frequencies so it's important to keep in mind you know what sort of orders of magnitude are we discussing when when we talk about those link losses just purely due to free free space path loss but now what about data rate so data rate according to Shannon we've got this bandwidth term up front with the w so the capacity is is proportioned to our bandwidth and then we've got the log 1 plus s in our term after that but the bandwidths too are just astounding when we think about what are the orders of magnitude of the bandwidths available and again i'll start with the u h f channels and we you know a tv channel is five megahertz at most they said they even subslice them to be less than that so five megahertz channel it's quite narrow for wi-fi and very challenging to get high data rates with such a narrow channel the wi-fi we're all using right now probably is is 802 11ax and five gearhertz and there are basic channels 20 megahertz but we can bond it up to 160 so we're we're getting an order of magnitude uh over the u h f uh bands and then in 60 gigahertz two two gigahertz channels now um so just you know keep gaining orders of magnitude in bandwidth it just is really important to wrap wrap our our minds around you know what the scale here is and then if we started even going higher 100 gigahertz up to tear hertz you know could we even be talking about tens of gigahertz or even 100 gigahertz scale uh channels which would just give us astounding capacities and now the the last aspect of spectrum i wanted to point you to is is a sensing resolution and there are lots of different sensing technologies i'll talk about some of them today but ultimately there's always a lambda in there somewhere limiting resolution wavelength and so just to remind you what those wavelengths are then we started meter scale with tv bands you know wi-fi were at centimeter scale millimeter wave 60 gigahertz five millimeters and then as we get up to 300 gigahertz exactly a millimeter and so as as we get to higher frequencies and lower wavelengths and we can expect that we will get better and better sensing resolution so the question could we have it all if we start going above 60 gigahertz above 100 gigahertz can we can we get the wide bandwidth higher resolution sensing can we use those gtgr terms that i didn't mention much up front but just said those are gains so if we can get good antenna gains and potentially we can get some of that uh those losses uh due to path loss back and and is that is that all it's about we just need to crank up the gains and and and then we're done and use all the things we've done in the past and so that's where i'd like to start today is is to say well what are the challenges if we wanted to use this very wide spectrum get the great sensing resolution is it just a matter of put some more gain up front and we're done so i'd like to talk about what what makes things hard and what are some of the research challenges what are some of the things we've done what's what's still open so the road ahead to you know high resolution sensing and and mobile and terabits per second and i want to talk about two kind of cautionary things too about how the control plane works and how we set everything up and then also i'll talk about security as well and so maybe to start um it's useful to ask a question uh how does a wi-fi work today at 60 gigahertz and and this gives us a a building block for for understanding how commercial systems can work what are what are the challenges what how does wi-fi at 60 gigahertz differ really fundamentally from from all the other versions uh subsets and so there's a there's a paper here um we wrote a few years ago with two of the leaders uh from intel claudio de silva and carlo squirtiero my student and i co-authored this paper with them about some of the features of wi-fi at 60 so maybe just to review this and so uh this briefly how how it works at 60 so um the the physical device is is an antenna array and it uses analog beam forming which i'm showing here with with variable weights so so variables steering phases in order to be able to steer the beams and um those weights are typically uh discrete ties and come from a set of fixed weights and then together those sets are called sector so you can you can think of a sector as a as a nice cone pointing in a particular direction as a illustrating there in reality i'll show you a picture of one and they're a little bit messier but nonetheless you can think of because the weights allow up to 128 sectors you could potentially be as as as narrow as three degrees beam width and the standard provides the mechanisms to how do you adapt the sectors how do you steer steer the beams which is um uh new for 60 gear hertz um for for uh phased array um beam steering and then that was able to allow an ad and then a y also added my mode capability so um so the steering alone isn't this my most space array but then as we have multiple rf chains um then we have the possibility for multiple streams and my moment that was introduced in 11 a y so the picture on the right showing doing a two rf chains a baseband and each of those independently steerable that's an example of the the capability that would be uh realized in in the 802 11 a y access point a client typically would not have uh mind the line so now uh beam training how does beam so now we've got to align these highly directional beams to get those transmit and receive gains that I mentioned in the earlier slides so how does that work in a standard um so the standard used a term called a sector level sweep which is a very in an intuitive way of of testing all the direction so we'll have just the the relatively simple view of of a sector just being a directive uh uh slice of the area around a transmitter and so the ap here tests all directions and um and the receiver is listening in what's called quasi omnidirectional so I drew a circle around the receiver here to say listen in all directions so uh the receiver identifies the the best one which in this case the the darkest green one um and then the process repeats in the other uh on the other side so now the the the client does its sector sweep and the access point is in a pseudo omni or quasi omni and and this process works they they find the both they both find the um uh the the best possible being that in this case the two two greens so it's so it's it's a successful um way of of finding the best beam pair um so this this is successful until there's some misalignment so here if we think about let's suppose the client rotates um if the client rotates then actually the ap is still fine and it's just the client that needs to re steer but if there's a translation as I'm showing then both the client and the access point need to re steer and so we've got to do it again so um and the other possibility is a blockage so maybe they steer towards each other and then now there's um now there's a person or some other object blocking it in the middle so they need to find a a a specular reflection off of the surface and create a non-line of site link so the standard does all of these things and um so what could possibly be wrong if the standard's able to already find the best uh beam pairs and it can already adapt to blockage and and and mobility then then aren't we just done here and um and the issue is is the control plane overhead and I want to give you a time scale here I'm not being specific with you know how many micro milliseconds everything's taking but this uh the the drawing is gives you a relative scale about how long does it take to do the sector sweeps for the transmitter and receiver versus how long does it take to transmit data and so you can see from the timeline below that actually the setup time far exceeds the uh uh frame transmission time and so that means there's an essence and outage that there could their time where the could be transmitting data but instead um it's um uh it's setting up and aligning links now you might ask well these standard guys just they they're not you know they need to get smarter and what what did they do wrong and if you and if you think about well why does it take so long each of these sectors has to be labeled so if the receiver wants to say I my favorite sector is green there needs to be some numbering on that and or some label and it needs to know who's doing a sector sweep so who's doing it which number is it and then that way they it can feed it back so what does labeling sound like it sounds like a header right so if you have a header and it's in essence a a a short frame and so the overhead is solely due to sending a physical layer preambles and and all those um uh packet headers or frame headers uh one at a time up to 128 so it takes a while to send 128 um different frames even even if they're short frames and this is getting worse so I told you I would show you a real beam uh from 60 gigahertz and that's one there on the left uh and so um but there's one on the right from 300 gigahertz beam and so you can see on the left you've got this kind of blob that you you can steer it around and the directivity is changing but as we get to higher and higher frequencies the directivity the directionality is getting higher and higher so what does that mean the search space grows so if if you look at the beam on the left you might think oh maybe we can have 10 or 20 of those and and select but if you look at the one on the right you'd probably think boy we need hundreds of those to to to get the best one um so uh so we're getting more possibilities the search space is growing and um and every time we retrain we have this outage that I mentioned and what that means is um the throughput ends up being far less than the physical layer rate that the physical layer rate we might have data rates uh gigabit per second scale hundreds of gigabits even yet if we have outages um then those start to really hinder what the ultimate throughput is so the question is can we can we rapidly re-steer the beams so I'd like to um talk about a a new device above 100 gigahertz uh called leaky wave antenna and uh a leaky wave antenna all as as I'll describe it and as I'll use it in our experiments as well um is a simple parallel plates that that um that maybe you use in in physics class for learning about wave guides and how waves propagate through uh through uh metal uh between metal plates and um and the interesting thing about leaky wave antenna is that it's got this opening in the slot uh so that's a slot shown on the on the lower left and what the leaky wave antenna does as the name suggests is that the the um the wave guide doesn't let the wave go straight through from left to right but but instead it leaks out of the slot and it leaks out with an angle that is related to the frequency and so um this is angular dispersion in which case the the higher input frequencies uh emitted a lower angle um so according to the equation on the right where c is speed of light b is the um the the distance between the two parallel plates and f is a frequency so the key number there is the frequency f determines the steering angle uh now I put new device in quotes because leaky wave antennas are really not that new they've been around since uh there's the patent application you can see 1940 uh and the patent itself uh highlighted the same feature that that I uh suggested um previously so they've been around a long time but what uh what really is new about them is that as we get wider bandwidth we get a much bigger range of f's that we can put in that denominator and with that bigger range we can have uh better uh control over the angle so for example we can use this property of angular dispersion to steer the beam that um that by changing the f from the denominator of that equation we change um the the angle that it steers and depending on on b then we can steer across a whole uh almost a 90 degree range 10 to 80 degrees um by changing the frequency so in this case changing it between about 150 gigahertz and and 800 gigahertz and so this has yielded the highest frequency beam steering mechanism demonstrated to date because it gets more and more challenging to steer using traditional methods like based arrays at higher frequencies above say 400 gigahertz um and it should be noted this is this is one antenna there's a it's a single leaky wave antenna there's no phased array there's no mymo and we have terahertz scale beam steering yet the question still remains where to steer where is the client how do we know um which direction to steer to which which which which is the best angle for the client and this is one of our our works um recently uh and we call it a terahertz rainbow and the idea is to excite the leaky wave antenna with a broad band pulse i.e. we put all frequencies in and so if we put all frequencies in then the leaky wave antenna separates those frequencies according to the angle so just analogous to the to a rainbow the lowest frequencies would be closest to broadside and the highest frequencies would be closest to parallel now what this does is it enables the uh uh instantaneous direction finding because remember the whole problem with the sector sweep is you have to label each angle say this is this is sector one this is sector two and so on and now here we have an inherent labeling of the frequency that if showing here in the picture the receiver is is located at yellow so therefore the with the single pulse and then the receiver is now localized with respect to the transmitter for the angle so with with with one shot so this gives us with the terahertz rainbow we have one shot location discovery and so with a single pulse sub nanosecond scale now all receivers i'm showing one here know their relative angle so the the the uh the method is remarkable for the for the things it doesn't it doesn't need the trial and error testing it doesn't need phase information it's a non-coherent method and therefore all the the setup required to to requires coherent communication signal processing and so on are not needed and there's no array signal processing so it's a very simple method to fast and simple so just to give you i want to show you how well it worked in the lab so um the um so far i've been showing you the the the equation for the peak emission angle on the left but actually it's not just one narrow band frequency but it emits over a range of frequencies and so you can characterize what that range is so there's some there's some analytical models that characterize how the electromagnetic wave uh irradiates as a function of the the slot width the plate the difference distance between the plates and all those parameters and then you get more of a spectrum um uh shown on uh on the left or on the right i'm sorry and then so we did an implementation in practice with uh in our in our lab with parallel plates and what we ended up doing the theory on the left uh is on the left and and the uh measurements on the right so what we ended up doing is creating a signature for for each angle so let me see if i can get my my pointer going um so the way to think of this is that if the receiver say here's a 30 degrees then what they'll see is not only the spectral peak around 300 gigahertz but they'll also see a fading off um uh as it goes down here but they'll they'll also be a higher mode um uh from the second mode uh t e wave and so you can see that the original model here was t e one but here we'll see two three and four modes and each angle then can have a signature so we can in one shot get the signature based on the the experimental characterization of the antenna um the theory just using the model alone works quite well too but we can refine that to get even better match so just show some results from the lab how well can we do angle of arrival in one shot um and so showing here measured angle on the x axis and the estimated angle on the y axis and you can see um with within the um the the core of the 20 to 60 range we've got extremely low errors and then it starts to tail off on the two edges and if you wonder why is it tailing off you can start to see well here we've got a lot of smearing out of of the signatures and then here the signatures are quite sharp so in the sweet spot of the range we're able to do uh highly accurate and of course we need to steer over more than uh 70 degrees as shown so we would have to have multiple of these and um to to get a full 360 coverage in any case so um that result that was uh with a for a PhD student of mine uh Yasami Gossum Porsche is now at um at Princeton and it appeared uh in there's a version of it in mobilecom and then also one in nature communications um and so I'd like to turn a little bit to sensing and um and radar and the basic idea of radar is is to send out a um a radio signal and look for for the received reflections and to use that in order to localize devices so it's it's still a localization problem but it's it's different in the sense that prior we had a transmitter and a receiver who are helping us align and here we we have a transmitter that's also acting as a receiver of its own um localization transmissions um so this um I'll just show you one one result on this which um really impressed me this is from the middleman lab at brown university and so um his idea was to use the same tear hurts rainbow uh method for radar and so it's the same thing if you send out a high broadband pulse through a leaky wave antenna then a rainbow emits and now the the reflection is going to come from a certain angle so the reflection will come back at a certain frequency i.e a certain color so that would say oh if I see something reflected red then it's close to broad side and you know violet you know closer to parallel and so the same principle and um and so the experiment i'm showing here is to take a a a small uh metal rod uh with 10 millimeter radius and move it along a triangle so the triangle shown in blue so blue is the ground truth and then um and then he's shooting uh tear hurts rainbows at the rod and then estimating where it is and the estimate is is in the red and to me the astounding thing is when you just look at the scale of this that's that's millimeters uh on the x and y axis and so we've got millimeter scale uh accuracy um or angular resolution of about a degree and again the the the thing that's amazing is all the things that this solution doesn't have there is no array processing right so you think about high resolution sensing that you're seeing in a lot of the uh research community's papers and and I mean eight antennas is kind of a starting point and you'd really love to have mass of mymo to get good resolution and this this is a single antenna there's there's one leaky wave antenna so truly astounding resolution um that we can realize by using this property of angular dispersion um another work um this work is by kaushik san gupta from princeton and so he um took the idea and uh and built a seamos circuit and we in our lab we were using a time domain system that was spanning a hundred gigahertz to tear hurts so that's quite a wide bandwidth to to use and so he not only did a circuit realization but also brought in a more practical constraint that well let's not assume we have nine hundred gigahertz available to do a localization and so he squeezed the the bandwidth required down to 40 gigahertz built it in seamos and demonstrated 3d localization using the same principle so i was also quite impressed with that result so um the the sensing and communication takeaway and then i'll turn to security for the remainder of the talk but the sensing and security takeaway is that the way it's being done right now at 60 gear hertz are uh phased arrays uh as as the main solution uh for uh steering and getting that transmit receive gain that we talked about early how important that is uh unique id for each beam trial and error testing um microsecond scale uh process and um if you want to do localization you can either use the trial and error or you can do array processing uh and then so with with a new device that has angular dispersion there's other ways to get angular dispersion from devices too but this is this is a exemplary one and so devices now now physics properties that that really haven't impacted us as researchers at lower frequencies that for angular dispersion it exists in other it exists in a phase array but but the the the impact is so minor that we can't really make good use of it for things like localization so it's something we typically ignore so we if your sub six gigahertz you probably want your whole career and never said the word angular uh words angular dispersion before so a single antenna and a spectral signature at each angle get all directions at once nanosecond scale and then now we can do radar with high resolution and and fast okay so i'd like to turn to security and ask about um security capabilities and threats as we go above uh 60 and above 100 gigahertz millimeter wave and a subterras and so i'll i'll just start with a with a one or two slide uh review of what what it is we're trying to do with security and there are many things uh for example like denial of service availability and and i won't be talking about those i'll only be talking about the confidentiality goal and that goal is we've got an eavesdropper in somewhere in the network and um the the goal of confidentiality is that if the eavesdropper intercepts the message then the bits will appear to be random so eve can no longer uh intercept can no longer interpret the bits that she she intercepts and only bob can decode the message and we assume that bob has the the secret key so in this scenario i've got alice's servers bob is the receivers and eve is sitting there wirelessly next to bob somewhere within radio range in the internet we have multi layer security so there's secure sockets layer ip sec um wi-fi has encryption as well so there's layer upon layer of encryption that i won't review today um but i would like to ask a question about why is there multi layer um and um the uh this this this cartoon that uh one of my collaborators made for me uh illustrates with the house and so we don't just put one block on a house but we might also have a security camera uh we might have a front gate could have a keypad could have a bucket of boiling water uh we can have all sorts of different layers in case one is breached right that's the point of multiple layers is that one could get breached so we want a backup system for other ones but why go to the physical layer why is this just a fourth layer and then are we really saying three will be broken we need a fourth but but from my point of view the wireless broadcast is the most vulnerable component this is where the signals are going over the air and that gives eve an excellent opportunity to intercept them you know versus on a fiber where it's extremely difficult for eve to even have access to the bits in order to attack um so at physical layer security the goal is that eve can't intercept the bits at all and it's a real new foundational layer of security so back to our analogy the reason i like this analogy is in in the in the one on the left we've got many many layers but the attacker there it has the ability to to try to pick all those locks right that's the attacker has all those locks in front of them it says let me throw whatever i can at this maybe i'll throw quantum computing at it right and then now if we succeed with physical layer security then the attacker can't even see the locks can't see the information can't see the signals and therefore the ability to to pick the locks is now gone how can you pick a lock if you if you can't even see it so that's the goal and i i'm the reason i'm so excited about physical layer security is is is not that you know three wasn't enough let's go for four but rather that let's let's close the entrance to the uh to the attacker all together now there's some hope that well if we have narrow beans did we just completely um thwart eavesdropping and so you know the vulnerability for eavesdropping in wireless is the broadcast so if i send out a message in all directions then the eavesdropper will also hear the message right but so there was some optimism at 60 gearhertz these are actual quotes i won't name names because i'll show they're not correct uh but the the the hope was that at 60 gearhertz it'll be covert so you don't overshoot the receiver or that the narrow beams will provide security because only alice and bob only bob can hear the beam that alice transmits and eve is too far away so she's not going to get it so uh securing a highly directional link so suppose alice and bob have a highly directional link and i want you to think about two scenarios wireless lane and rooftop so the the wireless land scenario is there's an access point bob's got a laptop or device we have a narrow beam are they immune from eavesdropping in the middle attack and what i'll convince you of is they're not and um the reason they're not is a strong adversary can now manipulate the electromagnetic waves through uh meta surfaces and so uh i'll call this the meta surface in the middle attack in which an adversary eve um places a meta surface between alice and bob and um uses the meta surface to allow alice and bob to keep communicating so if alice just or if eve just puts a big radio there and blocks it well then the alice bob link is gone and then they'll stop communicating so eve wants alice and bob to be able to keep communicating yet uh yet intercept and diffract some of the signal away towards herself so can she do this so some of the research questions can eve design a meta surface that's got to be a transmissive meta surface to allow the signal through to bob um what does she use for meta atoms uh how does she arrange them how does she build the surface and all of this we had a paper recently at y-sec and and some ongoing work as well um that i'll i'll i'll show you at the end so how does this work so if eve wants to design a meta surface for purposes of eavesdropping then she's going to target an angle theta towards herself so she'll put the meta surface for for ease of discussion she'll put a broadside between alice and bob and she's got an angle theta where she's hiding so she's hiding theta degrees away from from bob and she's going to choose a theta if not too small she'll get caught so she doesn't want theta to be a couple degrees she wants a a larger angle for herself to be away so the way she'll design the surface is um an anomalous diffraction uh via generalized snel's law and so um i highlighted the part of snel's law that's for generalized snel's law the the without the yellow part is is what you probably remember from high school physics and you'd say well we've got some index of refraction for the two mediums and an in going and an outgoing angle and you'd probably think well that's not going to help us at all the two mediums are air and we've got broadside incidents anyway and indeed yeah that part is is actually not that's not at all what eve can use but with generalized snel's law if she puts a phase gradient on the surface and that's phi x is the phase impact that she'll have at the surface and she can't just have a constant phase impact if she just delays the whole signal by you know 30 degrees it just goes right through a little delay so she's got to actually change the phase profile in order to um uh diffract it away and so that's the d5 dx and then c speed of light and half is the carrier frequency so that's eve's goal and so the way to think about this is if you like a material analogy uh with the original snel's law you can think of of of a lens for example and having the shape of the lens to to to impact the the the phase of the wave and and a re-steer it or focus it for example in the case of of other lenses um but now we're emulating that with a with a meta-surface and so we've still got to control the phase so how do you control the phase so we'll have a split ring resonator uh atom design and what this is it's a sub wavelength metallic structure and it's got a few parameters of interest the the radius the opening angle and the orientation angle and the great thing about this is by changing the all these three parameters we change the phase and so we've got to change the phase across the surface and so now we can select whatever phase we want on the surface and and therefore steer the beam so for example let's suppose you're um you're designing a particular atom in the surface and you say I'd like a a uh a phase shift of let's call it 50 degrees which is dark blue in the lower right then you'd be looking at a radius and on the uh if you wanted a phase of this color then you'd be looking up in this region you say okay my radius should be about this and my split angle should be about 130 so you say maybe radius of 250 um micrometers and and and a split angle of about 120 and that would get you the the phase that you wanted and you don't just want to repeat that atom over and over you actually want a distribution of atoms so showing here what eve is going to do so eve now has a desired data um and she's going to choose a phase gradient to yield the desired data so the way she does this is um uh she's going to choose different atoms and we'll arrange these in columns so we say one two three four five six seven eight uh atoms uh all arranged in columns and then that will repeat over and over for the entire surface um because we've got to have the same b5dx over the entire surface so that's going to be two pi over gamma so gamma is a distance here and that's going to give us the theta so I'm showing here a spatial period of eight meta atoms over six millimeters and these are the actual ones that we're going to use in our experiment and and these are these are experimental results of the surface and so what you can see is in red eve did get a successful two pi uh um uh distribution across these eight atoms and so that means she's got a d5dx of two pi over gamma so she successfully had an angle I'll I'll show you the next slide what her theta is going to be I didn't mention um amplitude she would actually like this blue to be flat um it's it's experimentally obviously not flat and we'll see how much that affects her ability um to uh to get a good signal at her location okay so now how do we implement these so um there's some different ways to implement uh meta surfaces and the traditional method is photolithography which is is wonderful and high resolution but um relatively high cost and and and slow and so the method that we used in in this work is a rapid and inexpensive fabrications as a printing method or in essence printing meta surfaces um uh with with hot stamping and so the reference on how to do this is below and uh it's a it's a wonderful method where you print the pattern on on paper so paper is transmissive at these frequencies so that's our transmissive substrate and uh paper print it using regular laser printer um and then feed a metal foil and the printed paper through a laminator and what that does is the heat and the pressure bond the metallic powder of of the metal foil with the toner from the laser printer and once those bond then you end up with a meta surface a printed meta surface a metallic printed meta surface shown on the rex is a wonderful technique for rapidly creating meta surfaces um so we we built that uh we implemented a meta surface and so here i'll show you an example experiment and um uh and so this is uh a tabletop setup so um one thing we don't have here is high transmit power so we've got a less than a microwatt over the entire terrarium so by having such low power we are limited to tabletop scales um we do have other platforms i'll show you in a moment for longer distance links but this one is limited to tabletop due to the low transmit power so eve prototypes a hot stamp meta surface in the middle um her target center frequencies 150 hertz and she she targeted 22 degrees and bob is broadside to the meta surface so let's see how well this works so first of all let's put eve 22 degrees away from bob and see without the meta surface what she's getting and this is showing what she's getting and the answer is she is getting noise and so she can't hear anything and now we place the meta surface and she has 20 to 40 db gain uh over uh a range of about 100 gigahertz remember her goal was 150 gigahertz center frequency and now she's got 20 plus db gain boost centered around there so eve we will call successful um uh at this um angle that she's targeted of 22 degrees but now we can ask the question about well what did bob get because if but if she blocked the link for bob then bob is going to discover it um or that that that something is is going on and uh or if it's if let's just say there's some big resonance if there's some big dip somewhere then bob's gonna see oh the signature in there of an attacker and i won't use it so this is the link without the meta surface in the middle and then when we put the meta surface in the middle there at bob bob is um surprisingly to us only impacted by several db so um so the impact was relatively minor and this would this several db is you know on the order of what one would expect with with really modest changes in the link it's not a that's not considered a very large change so that would be very hard for bob to um to discover um i mentioned too that there's a possibility of rooftop links and uh and so we're exploring the ability of drones to intercept rooftop links right now so we've already done it in the lab so um this is at josep torres lab at northeastern and um and we'll be doing a roof rooftop link there and so there's a drone off-shelf drone the drone is carrying a meta surface which is just sheet of paper not very not very heavy printed with some of those metallics split ring resonators and then there's bob and bob and so the the lab was successful and the the rooftop is is coming soon now if you know one question is can we do this dynamically uh can you dynamically change a meta surface so what if you need some move or for other purposes um can you dynamically do it and so there's some recent work that's not by me this is again kaushik's and gupta's group at princeton about how to create a dynamic split ring resonator so what you really want to do is just remember i say you want to open and close it and and rotate it and so how do you do that and so that was by uh using electronic actuation of a ring with some the ability to connect or disconnect different segments of the ring so you you digitally and dynamically control to to short or open portions of the ring and that in effect changes the the radius and orientation angle so this was some work again he implemented it in cmos around 300 gigahertz and uh and the the results were were extremely impressive and he even demonstrated some uh holography at those frequencies so we we need a much simpler capability for for the meta surface in the middle but of course with solely diffraction but of course you can use these once you have complete control of phase and gain now you can use these to realize much more sophisticated waveforms so the security takeaway is the high frequencies and high directivity provide new capabilities uh but also there are new threats and um and so the meta surface in the middle attack uh is is is a way of thinking about what are some of the new vulnerabilities we have even with highly directional links there's lots of open research questions programming the meta surface in the middle as i mentioned previously counter mechanisms how can alice and bob do better to try to detect devices in in their environment counter counter strategies actively securing links with the meta surface that that's right now we consider what if eve has the meta surface but of course we can give give a meta surface to alice and bob as well and that will will help them um so um just a couple concluding remarks and then i'll be happy to take questions some of the open challenges the components so the circuits and and uh making them low cost programmable energy efficient um sensing i i showed an example of sensing and where is the metal rod right but we'd like to sense a lot more in an environment in our environment and the thing i'd like to point out is uh at higher frequencies and trying to do sensing from an h matrix i think it's just very natural to go straight to a black box uh learning model but here we've got a really nice physical model as a starting point so i think that this will really be a new way of working at sensing um third uh we're not going to throw away below six gigahertz that that is we need those frequencies too so they'll all be fused in in the future and how to best do that uh network optimization management to harness meta surfaces that these these are coming um i showed one very specific example of using them but there's many much other work too about using them to increase capacity and get around blockages and so on so that's a challenge moving forward um and then new security threats uh and capabilities so um i'll i'll i'll i'll leave it here with uh you know just to remind you about the the we've got wide bandwidth uh directional links new devices and and and new security and i'll just end with some acknowledgments um the the phd students alumni at my group web page and the papers that i mentioned that we wrote are all there um collaborators um dan middleman at brown university joe septor net uh who is ian's uh a foreign phd student at northeastern university and kaushik san gupta at uh at princeton and i'd like to acknowledge the sponsors to arl nsf sisco and intel with that thank you very much i'll be very happy to take your questions thanks a lot adbert and uh is there any question from the audience please uh type here in the qa section uh i hope we can get some questions in the meantime i can start to ask adbert uh because there's nothing now so far it looks like everything is clearly explained uh yeah so i would uh focus more on the second part on your uh meta surfaces as you know uh meta surfaces or intelligent surfaces or inconvenient intelligent surfaces are beaten to death the last say six years and we did research almost 10 years with the european community uh it's called wiser surf project with uh uh crete fourth research center leading it we worked like four or five years and there are zillions of papers in that wiser surf dot eu so uh and then there are many hundreds of other papers so it looks like you are trying to introduce these meta surfaces as eavesdropper kind of like uh you know denial of service attackers or whatever uh so my my question is kind of like you're getting it from another perspective uh i mean these meta surfaces are helping actually to avoid security breaches you know we showed it and others showed also that when you design meta surfaces you can uh take care of the security attacks and now you are doing the other way around that you say meta surfaces can help you to do security attacks so is it uh straightforward so i i don't get the point here yeah i um for sure that um that meta surfaces can be used to uh to help secure a network um the the the reason that we're exploring this direction is we always want to consider the strongest possible adversary and so a strong adversary as we introduce new technologies on on our side you know when alice and bob say let's use meta surfaces um we have to also consider that the adversary will have one so i i would say it's it's a consideration of the next generation strong stronger threat when that threat uh is equipped with all of these uh advances that you mentioned because it is highly active field and there are many meta surface designs and so we have to think adversaries will also be using those advances so you are but the issue is the following so if the meta surfaces are deployed between bob and alice and then you say eve can come and deploy her own meta surface and try to do eavesdropping but then those meta surfaces there between bob and alice are not strong enough because they cannot avoid the security breaches right so and then also where will you place the eaves uh uh meta surface that's also an important subject you know like how will you try to get that information right but maybe you're still continuing the research there are other problems i mean questions uh adrid let me uh read them to you sabar kamoshi is asking do you see any future for over 60 gigahertz and terrors to compete with uh come on anyhow i have to read with existing under 60 gigahertz for indoor coverage i mean solving indoor ultra dense networks i'll let you answer it okay um so i'll answer this one and i'll also just one last comment even about knowing where alice and bob are in in in the scenario so for sure eve has to localize them i think that the scenario where that's the biggest threat is um a rooftop backhaul because then it's very easy for eve to know where the transmitter and receiver are because they are they are highly visible and they're non-mobile i think in the mobile case then eve for sure has her hands full she can easily localize the access point but but then she would have to also localize bob and then place the device i would be kind of like putting a bug right if it was an audio bug and somebody wants a bug put an audio bug in your office right then they would have to you know get it get it in the right spot um so for sure challenges there um but back to the wireless LAN question um for i we i had the bullet about jointly using different frequencies um for a really specific reason that we've always got to have the sub six gigahertz for coverage because we can't have a link outage you know if you close the door to your office and the access point is outside the office and and and it can't penetrate through through the door for example so we will always have both i think the way that it's the the ic wifi using higher frequencies is through the the the type of multi-link operation that's that's just now being standardized and and also the channel aggregation and so on where there's going to be a base channel below six gigahertz that keeps you always connected and then opportunistically that if there's also a two gigahertz 60 gigahertz uh two gigahertz channel at 60 gigahertz or higher to to terahertz then those will be used for performance uh enhancements but you'll always remain uh connected at lower frequencies that that's that's the way i would see industry in evolving um over say say with wi-fi eight that that would be my crystal ball prediction for wi-fi eight and uh there is one more talk but here's a comment uh you also mentioned the actually i remember you were working on super wi-fi i remember like 10 15 years ago and then it died down right because it's supposed to do like these uh secondary users of tv bands like super wi-fi but it never made it right i mean are are you following it still what's going on there are they having some businesses and super wi-fi yeah so the the the the challenge with super wi-fi uh has to me has always been the the us lack of channel availability in urban areas and the um the the very difficult business case in rural areas so if you're a startup saying i'm going to ignore all the urban areas and only serve rural then that's that's a much more challenging environment to work on nonetheless it is still a a a rural broadband technology it it's but but the market because it's rural that puts a big cap on you know where those markets will be so that that's that's why you know you're not hearing about it every day is the next big 6g thing because it can't it it's it's a rural largely rural only at least in the us because we have so many channels let me add something my personal experience uh last or before the pandemic uh georgia state called me and uh i was really surprised even in georgia we have rural areas believe it or not we do not have any coverage the only coverage they have is through the cellular technology really you cannot believe that and it's really expensive right to get internet services through the cellular phones and they were discussing what can they do to you know provide the services to these rural areas and then they made you know we discussed in a lot of things and then they said you know maybe it will be too expensive for the state to go to those rural areas and provide services and there will be no company to provide services so they're really like she can act problem they're stuck you know i feel sorry this is a us even you know we're not talking about other countries so anyhow if we don't want to have that dialogue here let's go to the next question by hong ji guo and uh uh for the meta atoms will the seamos switches create non-linear effects at terahertz it seems not trivial to add the seamos switches what are the challenges in the implementation so for sure it's it's not trivial um and and you know this is uh kaushik sin gupta's you know claim to fame is how do you build seamos uh meta atoms that are switchable uh not only just switchable at all but switchable at frequency is sufficiently high to be valuable for communication so i won't attempt to say what uh what the the seamos uh circuit design challenges are because that's so far out of my uh my area of expertise but i i would um let's see if i can put that slide back up because i i would suggest to uh uh to look at that slide um or that paper there it is uh nature electronics and um and and the the design specs and the design challenges are all in that paper okay are there any questions i think that's about it or so i don't want to take more time of yours or uh one more question maybe while i'm asking you some others may get in for these leaky wave antennas in the first part of your talk uh about the performance you know in the higher frequency bands we have the distance problem so uh can your antenna design help to combat the uh distance problem how far can you go i assume are you still indoors right and how far can you go up yeah so for sure we're still indoors and we um we are just now starting to do some some longer links so far we've relied on our collaborators such as uh joseph jornette for the for the longer links and and those are typically uh in in the experiments um non-steerable links so that still is something we have to experiment with about the longer distance um steerable links with leaky wave antennas so it's ongoing work as we move from the you know the tabletop where we're low power and we have a lot of control a lot of capabilities to to experiment with new antenna designs but then as we go to the longer distances and longer range there's some new challenges we do not have for example steerable phase arrays uh at these higher frequencies currently okay thank you edward and there are no more questions alessia i would like to ask you uh to get in and uh go to your wisdom corner questions thanks a lot again edward okay thanks again i appreciate it thank you very much ian for moderating this question this session of q and a and thank you to the speaker for your very informative and comprehensive presentation so now we move to the uh wisdom corner live life lessons which is based upon the idea to give a unique special angle to this webinar series adding a personal touch so we we have during the the wisdom corner we have uh successful researchers like a professor like me today to guide young students in the current ict research field so i would start with with my first question i have a couple um so which strengths you believe at which capabilities students and young scholars should be most focused on developing and how do you think they could accomplish in uh this well i would say um for for future uh network design uh interdisciplinary knowledge and collaboration is extremely helpful and um and so we we as you know systems researchers we experimentalists we build build systems and in the past it might have been possible to do it all within your lab without an interdisciplinary uh collaborator and in fact some of our older products we did they were solely uh myself postdocs my own students um but i'm finding that that's changing especially as we get to uh higher frequencies and um and so i think that the examples i gave my talk uh including uh kashrik singh bupta who's a circuits expert uh joseph more on the communication side dan middleman as a physicist and and they they bring new knowledge that um uh and expertise that that i don't have in my group and so i would say the advice is is is to uh learn to work with other disciplines uh learn to learn enough about their areas so that you can effectively communicate and collaborate with them because in some cases you might find there seems to be a a little bit of a uh a language barrier in terms of the tech lingo that that uh that we use and how we communicate um but it's it's super valuable and i would say related to that too is is making sure you learn enough about their areas that you don't just delegate and say okay you're the physicist i'm not going to learn anything about how these you know meta add-ons really work you know you just get them working and then i'll make a network out of it but to learn enough about what their capabilities and limitations are so that we can um uh co co design uh what uh what they what they might achieve thank you thanks a lot and uh let's go more into the some technical details in which fields specifically at which topics you would recommend uh young students to study uh yeah so i i'm an experiment so when i get excited about an area it's something that maybe uh is is emerging as a new um uh as a new capability experimentally and so like for example ian mentioned before about how meta surfaces have been an active topic for uh for a long time and um and uh and that's that's definitely true but it's also true that we're just now starting to see the capabilities of of um like the circuit CMOS circuit that i just showed you where those are just now getting to the point where they could be viable in in a in a larger system in a operational system and so that time where you see the cusp of this is this is right before we could really start to put it into standards and and working products i find the most exciting so that's why i would say in terms of topics um in the one i had in the title of my talk is you know sensing and communication and and networking and security all at these higher frequencies because i think well we will see new devices like the leaky wave antenna this is i'm thinking another great example where it's a device that just forces you to change the way you think about networking that rather than adapt our current methods and tweak a few things to make them go faster but to maybe rethink about how we perform some basic uh basic functions on the network right thank you and let's go maybe with the last question into some more personal uh anecdote tell us one of the uh of the most tangible contributions that um had had a strong impact on your life on other slide that you're most proud of yeah that i would have to point to the technology for all network that ian mentioned during the introduction that when he visited in in 2007 and and i took him to to see the network but what what that was is um as as the name suggests technology for all it was it was a network uh for an underserved community in houston and um and you know the story of that was in briefly the short version is in in uh 2003 we had a large grant from nsf to design next generation wireless systems made the front page of the houston chronicle and the founder of the organization technology for all uh read about our project and and he he contacted me reached out to me and said that we we have this neighborhood where we have a program to give uh pcs donated by industry to low income communities but then once and we train them so this technology and training so learning to use a pcs um and then the then they're stuck because then they can't get internet access and then without internet access all that training and the use of the pc is just not valuable and they can't do it because they couldn't afford internet and so he asked he challenged me can you take the wireless technology you're developing and and and i give it a trial in this neighborhood and and i'm an experimentalist but there's a big difference between being a lab experimentalist and being an in the field operational network experimentalist so it was a massive massive challenge that was about the time i just got tenure so i thought okay this is this is this is the time to take on such big uh big big risky very high risk uh project so we built a completely custom mesh network and so um meshes when wi-fi access points multi hop amongst each other uh and we blanketed the neighborhood with uh with multiple mesh access points and wi-fi coverage and provided free wi-fi internet to the to the neighborhood with our custom nodes because there were no commercial nodes at the time um and then uh since then a whole commercial industry has has emerged and it's the technology standardized and commercialized mesh is a multi-billion dollar industry today that from homes to outdoor areas use the multi-hopping technology but so today it's still running it's a it's it's an operational network using um using commercial mesh today but the but you know back to your question about that was the first time that we were really meeting and interacting with the end users of the technology so you know for example the things i was presented today you know there's so many different things between me and the ultimate user of these technologies this first got to be standardized commercialized and then maybe down the line there'll be users and this was we we built it we designed it we wrote the software built the hardware and and literally put it in people's home so we met the people and said yeah here's a device and this should give you free internet and they were super thankful super happy they would call us when it stopped working right because we would have outages and we would get phone calls and so on so it was really a very unique experience but also rewarding because we got to see the impact of the research directly and immediately and and meet the people so that was a really valuable experience wow very interesting thank you this really answers my question a very strong impact when you see uh that impact on people that's that's i think the most um you're grateful for uh thank you so much and i don't know ian if you would like uh to ask uh some questions to our speaker or we uh we conclude this session i think we are done and i would like to thank edward again for accepting our invitation and delivering this outstanding talk and uh i hope to see you edward in person someday in 2023 hopefully no no other pandemic will show up right and uh thank you alessia for thank you so much thank you both and uh i i ask uh the audience please submit your papers to our journal and moreover soon we will have another great speaker and our webinar will continue we have like three more lined up so we look forward to seeing you in the next webinar again thanks a lot edward and latest i assume that we'll see you in istanbul for balkan come right sounds great yeah looking forward to travel yeah opening back up absolutely okay thank you so much again thank you bye bye