Home » Podcast Episodes »

Biometric Authentication – Alexey Khitrov, ID R&D – Voice Tech Podcast ep.024

Alexey Khitrov Idrnd

Episode description

Alexey Khitrov is President at ID R&D, who increase the security of interactions with digital products such as conversational interfaces using voice authentication and behavioural biometrics. Their products add hands-free, zero-authentication to any voice enabled device, heralding a new era in frictionless, ubiquitous computing services.

We discuss how voice biometrics move us towards a future where we communicate with technology just like we communicate with our friends. Alexey gives us an overview on the state of voice authentication today, who’s using it, and its limitations. We also cover the technical features of voice authentication, such as modal enrichment for continuous enhancement of the authentication accuracy. We hear about the design challenges faced when introducing zero-authentication to your application, and get advice for implementing this technology into your own voice apps.

To wrap things up, we look ahead to the voice technology features we are likely to see emerge in 2019 and beyond, and tackle the risks around data privacy, security, and the evolving arms race between hackers and software security firms. This episode is one the most interesting conversations I’ve had doing this podcast and it’s a real pleasure to be able to share it with you.

Links from the show:

Episode transcript

Powered by Google Cloud Speech-to-Text

welcome to the voicetechpodcast.com Robinson in conversation with the world-leading voice technology experts discover the latest products tools and techniques and learn to build the voice app to the Future so it’s easier to use it provides more functionality is more secure amended the list goes on

hello and welcome back to another episode of The Voice tech podcast my name is Carl Robinson and today’s episode is entitled biometric authentication your have a conversation with Alexa khitrov president at ID R&D who developed authentication technologies that increase the security of interactions with digital products such as conversation interfaces this includes voice authentication keystroke pattern identification behavioral biometrics continuous multimodal verification and are more besides all that does is it enables hands-free xero authentication to any voice enabled device is ready Herald a New Era in frictionless ubiquitous computing services is super cool Star Trek level technology that will eventually hopefully banish login passwords to history books so you’ll learn all about how voice biometrics move us towards a future where we communicate with technology just like we communicate with our friends alexey give us an overview of the state of voice authentication

who’s using it and what its limitations are what are the barriers to adoption and then we do a Live demo of the ID R&D serverless authentication alexey then gives us a bit more details about the technical features of voice voice authentication which is the modal enrichment for continuous enhancement of the Year authentication accuracy we discuss some of the design challenges faced when are introducing the authentication to your application and alexey shares his advice for implementing this technology into your invoice apps then we discuss the voice technology features that will likely to see a merging their 2019 and beyond we discuss the risks and potential risks around data privacy security and also the evolving arms race between the hackers and these software security firms it’s definitely one of the most interesting conversations I’ve had doing this podcast and it’s a real pleasure to be able to share it with you today ok so a quick update last episode I mentioned a couple of ticket giveaways

I’m the winners of the jackpot Super Bock conference 2019 tickets are Andrea Daniel ask you from Facebook and Jeremy Wilkin from VMware is also host of the design for voice podcast which is checkout to congratulation spicy enjoy the conference if you didn’t win tickets be still want to attend then it’s all happening on April 2nd in San Francisco California the link to register and is in the show notes and you can get 25% off with the code voicetech case we got a new sponsor that hexnub active.com quick disclaimer he is my little brother but I’m very very pleased to announce that the services the office because it says highly related eczema bar makers of robotic toy accessories are basically they enhance your robotic toys and give them or personality through hardware customisation to the trend these days is all about voice enabled robotic toys like the anki vector for instance and they are great both of the toy and because he wouldn’t want

talking robot but also as a means to learn and teach programming and stem skills to the Next Generation so if you’ve purchased one of these robot so I thinking about it or use these these robots in night in your educational establishment and then you can purchase a great range of accessories at Hexham up.com they even offer free 3D printing blueprints you can actually make your own accessories for free for your talking robots that was super cool you can check that out xm.com if you want to sponsor the podcast and get your brand feature on website the weekly newsletter and the clubhouse itself then you just have to head over to voicetechpodcast.com / donate what are you can quickly and easily become a champion sponsor to this episode is supported by dabble lab provide free video tutorials and code templates to help you learn to build your own voice apps at YouTube.com slash double lab will find over 150 video tutorials for both beginners and experts containing step-by-step instructions for how to build apps for

Alexa Google assistant Microsoft Cortana jovo Bixby and more as well as using industry-leading tools that just William bespoken all the free templates are at skilled hand plates.com and all the tutorial videos are at YouTube.com slash DAB so if you’re a voice that developer or you’re thinking about coming one you going to want to check out this amazing Resource and take your skills to the next level ok so now it’s my pleasure to bring me today’s guest alexey key Trove I’m on the line with alexey ke 12 the president at ID R&D alexey welcome to the show so much it’s great to have you on the show I have to say you’ve got a really cool products and after the thing I know this first one was while so excited to speak to you actually as that of all the technology products that I’ve come across recently this is one of the coolest isn’t this the one that sets my imagination going out the most ID R&D just briefly is

you guys develop authentication technologies the increase the security of interactions with digital products such as conversational interfaces that the strength of the website and apparently have developed a portfolio biometric technologies including behavioural biometrics and also provide R&D services for customers looking for an edge skin be fantastic to hear all about that and can you give us a bit more detail than on are on that on that brief intro how what is it that what does it really that you guys build and a who your who your clients are excited to share more information about the company in the end of knowledge so I didn’t he is the name would suggest is the company that is focused on one or fundamental science support of the biometric centre AE and application of the decor of the company is a group of ambitious scientist.

we believe that the deck and I’m now in improve the security of an occasion or a while enhancing the user experience and the way that we see you know what did the natural progression of the year or the user experience we see we will go round and see how we communicate with our in our friends and family so when you meeting your friends you’re very early start with providing you your username and password so you know who and exactly are we working with and we think that that’s exactly where they were the technology is going so so our goal is to build secure Enough by metrik capability that will it allow this is absolute seamless user experience when it comes to altercation are specifically in the conversation interface where it were convenient series 1 of the

or benefits of this technology is so so so so so our goal is to enhance the user experience to the point when there is no user experience whatsoever that is really the key right this is that frictionless experience where you can just walk up to a voice interface and start using it because it knows who you are is the key technologies about authentication so it knows who you are just want to sound of your voice and so you can access all the personal services from the moment you start talking and this distinction year did we were talking about allocation Ian security it’s important that the motor out indication technologies AO3 not one but two questions so it’s not just about is this the right person it’s also is this a personal and so because you know if you

Eastbourne the application system with a photo or with a recording of The Voice or with the computer generated speech than you know from the security standpoint it pretty useless piece of Georgia so are we going to get onto the security stuff in detail I think a little bit later what what kind of clients that do you typically serve like who uses the this, product service or working with the three types of customers one is end users and for us and we defined in uses of the companies who are The Enterprise customers that degrade or allegation capabilities within their products within 30 interfaces and we’re talking about banks Trust companies financial institutions and healthcare providers in someone also work with the album with the biometric in the security integrators and it finally we work with our large tech companies that are integrated technologies in winter level

there are chips or devices biometric integrated what’s that what kind of companies load Security Services are to the arm typically to the e to the Enterprise customers so they work are so so Francis wework most it with voice biometrics with voice overs indication they build platforms that can a human can utilise multiple modalities for multiple users use cases and interfaces in out shake your back ends in someone and we and we provide an automatic capabilities Court by magic engines that they are can I plug in the UK platforms that provide content service to the customer ok see you that you guys focus on the technology and then they perform the integration work with the with the end customers right and let’s talk about some of the products you offer then I had a look on the website there’s that the one in the

tickling the stood out was there safe chat 2.0 the world’s first zero effort continuous multimodal verification product for chatbots and virtual assistants to save chat to combines the power of voice face and behavioural Dynamics to authenticate a user without requiring a login and password details a bit more about that how does that work safe check is a Holy Grail for for for altercation because it delivers that are good friend friend the family use experience that we talked in the beginning all the conversation that the user can communicate with a virtual assistant assistant be through through voice input or Type II error code and conduct transactions or I asked questions about another state of the account or as some sofa

NFU the right user then if the happy path then it just a normal conversation what’s the balance on my account here’s the the app which show the balance can I transfer $100 in oil for my rent yes please you know can I find out you know how much I spend on milk last night but if somebody else wood in a wood crap you’re ready your device and would to anybody asks for information or to conduct a transaction that interaction with would immediately stop the key ingredient Heroes that workwear multiplying I can multiple biometric modalities at the same time through natural user interface or lack of the interface so you will be providing input with voice biometrics we will confirm

is the pickup the device and is immediately start divulging potentially sensitive information and discussing their bank statements etc and making transactions and for users who are currently are used to to some level of security before they start doing these things it must be quite unnerving it must be here is a whole new set of habits to learn or I still have us to unlearn really what have you experienced in your in your user testing that what’s been a reaction from users and have you managed to get them get them over the hurdles Kerry thought at first but once the user gets understanding technology and get to get used to this um free communication between the user and EE ie Enterprise there is no coming back this story about something completely

connective tissue do biometric world with which shows the importance of patient experience you know if you give me your ears back when blow Buster was one of the biggest retail chains in the United States they had a chance to buy Netflix and you started it was providing this much of a frictionless experience when it comes to renting videos and they conducted a vocal group and the most people in the focus group said well we actually really enjoying in getting off the couch and going to store in can chickens eat works 2200 peak amps eBay movies of course it turned out that once the market realise the b’s and functionality of nature of the experience was Netflix or has devised an internal network is one of the oldest company in the world

what’s the difference experience of people star salon x 202 estimate how much they’re driven by this is there of the simplified user experience uniting the product safety at the unique in a way that they do leave the door delivering this district nurse experience while providing the Enterprise with the multi-layered multi model continues biometric application experience in real time I can definitely see that I still think that for some sense of transactions they might need to be some kind of indication that you are talking on a secure channel because speaking speaking to a smart speaker you really have very little idea of you know what’s listening and it that there’s not much reassurance coming back with no icon on the screen or anything telling you that you know the pad like that says this is secure or whatever you know even if it’s just some icon that you don’t really understand it’s still got some some level of reassurance so I’m not sure what they’re that you iPad him around that is butter is interesting question it anyway they say

Jet2 it says I type or speak to your request for a chat box you handle typing as well to save chat in Canadian the way that people like to communicate with the with check and stairways to do it one is a more traditional three typing and one in working merchants Cowboys and we cover boys with her with biometric identity user base in just a few seconds of speech which is quite a new because traditionally takes to 10 seconds of speech but develop our own behavior by behavioral biometric capabilities and this capability allows us to extract enough you need biometric data anybody you divide the user

arthro throughout the transaction or through and we’ll look at the way that the person interacts with the device the way the person types and then say one and a few looking at the keystroke Dynamics that the actual pattern than tapping the keys and I also measuring other other parts of the phone as well using all the sensors on the phone at the gyroscope and interesting in this technology if you can I go back on here and history goes back to the Second World War 1 you know when both sides birthday on July forces in Germany and they realise then they ad3 really operators had their own wall decal signature Alice Tyler Moore using the morse code and then I’ll do the endless number of detective stories and you know and then x

Sky movies about how and one side with kitchen radio operator then make that person again then try to feed the wrong information to the other side and then that aside will recognise that the signature is wrong so now where does wee-wee understandably getting in a wrong information we’re going to see them back sauce that’s not stopping us now so so so the core idea is that the way we type the way we we will do the input information for typing through the weather with all the devices is quite unique and I’m going out with you with the introduction of the I keep getting much better understanding might better sense of who is typing who is doing the improved information and then we were also doing any other will always try to combine that with somebody

modalities as some additional data points from the device we also noticed the typical woman when people type they hold the phone in a way that I can look at the phone and therefore we can utilise the artificial for you can see as well so so so so basically trying to the wood trying to get as much information much is my due date is possible through the normally contraction bb20 using the device can you use it for the biometric analysis and Alliances how much data comes off of a mobile phone enabled on smart speakers which is got a mic on it and get your carrying around a mobile phone with just got 101 different sensors that measure mean from every possible angle in never worried about that but anyway yeah just the last couple of things the products I see you’ve got there and ID squared it says that the user simply typed in a log in a password and then I dsquared validates the user by how the user types in the login and passwords you do you do provide that simply for the login password authentication as well

and you can capture the user space at the same time and then there’s ID behave which is continuous continuous verification of a weird Musa in the background as a method to strengthen security other that was a really interesting because I’ve not come across that you know continuous verification method but you can imagine someone logging into a website using it before walking away and then somebody else sitting down and you know clicking away and at that point you want to kick him out because they’re not using it in the same way that the can I use case have I got that right now look at the moat Wrotham a more traditional use experience of fob username and password the biggest problem with the username and password is that if I know you are using my password on your computer compromise wagon login AED into your account however what I just called that is that we had the by the magnetic layer to the oven existing use Xperia

so we apply our behavioral biometrics to the way that you type in your you again we will also applied official recognition with live Nissan L200. And as a result we’re turning the existing a traditional username and password into the multifactor me tomorrow and application Experience episode episode the same where we are the children of changing it but in addition to the traditional knowledge based authentication and that’s exactly what would the password is already a biometric component so so sorry I didn’t behaviour with spatial variation that runs simultaneous to query with Winford of the older you ordering something and something you know about something you know

but really the face and the way that you hold the phone exaggerated something that you have it something that you carry with you all the time and the combination of those do it is much much stronger than and then anyone on his own I will go further so take these essential workwear how many years is that what you know it is the password but you have no idea can be your device and then another question is who you are and that’s the biometric security continuous verification that something could you expand on that because that is interesting absolutely so fucking teens verification I think it’s it’s extremely important concert because it’s because traditionally when we think about the application biometrics as something I had did the beginning of the session so we were only gonna garden the door but we’re not really looking at what’s happening inside once you’re getting an endothermic reaction

how to keep on confirming identity of the user throughout the interaction between the user and the antidepressant do in the way that doesn’t affect the user experience so it really helps from the securities DST and only see if you can guard against in Oldmeldrum animal attack on someone but it also helps with audio experience it also works perfect sense when calling my bank and I speak to a person they can immediately tell if I switch you know somebody like writing to The Liner start talking on my behalf but when talking to another and automatic air agent as not the case you can have to continuously verify it’s the same person talking all through this all through the interaction and I seem this is it is written on the side with regards to a web user but I seem to have this information is part of safe chat as well when we talking to her complaint address

hey I don’t second time golden state Warriors last time golden state Warriors ok enrollment is successfully completed prices just processing takes a few seconds to extract all the the features from my voice so we can authenticate me ok it’s done I like to know we’re going to move to the verification step and for that I’ve got the house of a lovely assistant my girlfriend Veronique so I’m going to try and verify my voice enter my username again ok verification golden state Warriors recording a successfully completed processing will I get into the system verification successful anti-spoofing successful anti-spoofing score 98% music

be confident that is a human speaking verification score 100% well that’s pretty accurate right so let’s try the other case now where my my sister is going to try and break into the system using the same passphrase but considerably different voices your hair ok so let’s let’s try it now with an attackers voice recording successfully completed processing verification failed and his two things successful and he speaking score 62% verification scores zero so it verified that you are a human or there less human the me and verifications 0ur you did a very loud impression you didn’t get into the system so the data is safe because he didn’t get in thank you very much for an egg alright so we’ve just had it in Action it works perfectly of course and obviously there’s a huge number of advantages that come to mind

sophiatown do you like this as we’ve already mentioned that it’s hands-free and it’s a frictionless is it it’s absolutely zero friction you just walk up to a device and start using it and it indicates you as you’re using it so there’s no training involved and is anybody else can use it means the elderly children people in foreign languages it really makes no difference who you know who are we doing with they don’t need anything to learn anything new because there’s nothing to do and that to me is just an incredible experience

voice chops Tuesday is a weekly newsletter to help you build better voice apps whether you’re looking for research chart depth charts or design chops there’s something in there for everyone just go to voicetechpodcast.com / newsletter and look forward to your Tuesdays another advisor comes to mind is that at the moment all of these smart speakers in the voice interfaces are linked to a particular person who’s logged in through the RAF or whatever out as soon as you have this implemented everywhere Mrs is there a friction authentication every single Mike in the world is suddenly at you available to you so you don’t have even have a device on you can just walk up to any any smart speaker or any voice enabled technology in to start using immediately and is accredited with the promise of the water quality IG8 so you can use this channel to conduct your all the test day

Fiat Fiat Fiat you might happen that’s my voice is becoming it’s actually the new interface also if you think about it we we used to do you know in 20 years ago we did everything on our computers in Aberdeen have a laptop back then but when I proper computers in about 10 years ago you know smartphones came into her life then or sudden it was all about apps in the earphones but now isn’t switching to the new experience to the voice experience when you know when you go home when you tell your device you know to turn the sauna turn this off when you know ladies musical play that music in in that is the beginning of the penetration of the older conversational interfaces into the everyday life is actually happening much faster than the penetration of the smartphones

general public 10 years ago so as soon as it is serious it’s amazing isn’t it incredible that be anyone that was doubting that you know that this is that this this is the future because it isn’t here right now you know that is not going to happen you know until it suddenly upon our school with that come from but it’s a smart Home device enabled we just seen the one of the large companies voice enabled toilet Nike Just of Wallace enabled shoes so it’s ok we’ll be coming OK Google Voice enabled and the voice biometric to you can bring actually one is the security so you can you to be able to walk into the interior fridge and say all in all I found out of milk in your eye I need to order some Bill

and if we should confirm that is your voice and European place an order in the natural experience of communication with your devices I mean we’re so used to you not talking the screens of my smart phones are typing on a laptop you think about do you know what’s more natural you know the natural is what we’ve been the humans doing for centuries and that is talking to each other in this is exactly what the F is going now as far as our interaction with other world of technology so so yes you can definitely work if your fridge and the end I ask for more milk but there is one more aspect and that is the aspect of not the security but the person is patient or services they mentioned when you get it to your car and see you now turn the radio on and your favourite

radio station turns on but then your girlfriend jumps into the current system during the radio on and then her favourite Rodrigo radiating plate so changes to know she’s already and it starts play something that knows you both like it’s all isn’t so much a security that were talking about your data security in a TV but which year was Reading Station you’re playing it’s not a secret transaction with occasional eating disorder and S Voice becoming interface is person vacation becomes more more more more important voice biometrics can provide the stability saw the we see are there in the absolute flooring demand for with my messages from the standpoint of security and enabling

mobile Money transactions through The Voice channel or sharing some sensitive data through the week channel also just for the convenience of our own be ugly will be able to the personalised AUD users without again without any specific apps from the user absolutely has is easier to use it provides more functionality is more secure I mean that the list goes on and what why is everybody using it right now what’s what’s what’s in the way what are the barriers to adoption and what needs to change for this to be everywhere I think Theresa traditionally divisible by Matrix was it was mostly based around the text dependencies you when there is a particular phrase text independent capabilities

until very recently required a longer voice samples over the age of 10 seconds at least of species 526 sentences that pretty long I didn’t seconds is 5 to 6 sentences authentication he has got a long pass phrase Ronaldinho interaction before you know it’s reliability especially around the X factors allows to do the context independent voice altercation be peace on just one to Toulouse to 2 seconds of speech in this is only changes in ODEON the hue the usability of this technology in the conversational interface because you no longer need to wait way to the conversation come from dead ntw3 you can do pre-match Rider waiting at the beginning of any conversations even know you’ve done it and it’s done

so I realise the wake up word as a dect the panel component and then you you can buy it with a text independent third component of the Fall on? So so that are not allowed to bring the on the accuracy at 2 required level 3 much right away in a security so so so so so we’re bringing in a DVD user experience through the surname they require others when it comes to the security so the two-component one is the in the accuracy of all of the word biometrics nothing that we’ve seen some significant progress there are all the order last few years through the introduction of the API based but we’re also seeing the question of interest aliveness or work.at spoofing protection being addressed that’s the question

that’s all that equips Odessa II by Metric worksheet not just a is this the right person is it a person it’ll and is the answer to those questions and I or is the ou ou sign that there’s a different question answer to this machine any messages for for both of these both of these problems in tandem they must work now well I in real time to provide the positive response to both questions in there and if it’s going to do then we can you know providing the information connections actions and sounds but now you know it’s perfect idea and you are a lot of attention specifically on the enter key capabilities in Wigan successfully distinguish between live voice in a recording overlay voice in the Stables live voice and the

computer generated to be any computer generated speech is getting better and better so so this is a very can I insure an ongoing battle for the biometric industry synthesised voice is getting better all the time so it sounds like there’s going to be some kind of AI Arms Race where you’re using AI to detect and they’re using a lighter to full auto circumvent that the security procedures in place it is that is that what you envisage on this going to be a bit like her computer viruses or it’s just a constant battle to beat the inner the attackers and one in the original species becoming better and better and the requirements of the target voice is getting smaller and smaller anomaly scan A&E unit meaning of hours of studio recording with target voice to create in our computer and change generated engine there was people like that person and last year I mean

allowed the functionality that can dye it should be seen with 20s minister of off-target speech and recently by doing China said that although they can do this and one minute speech so it’s one minute is other now I have no idea was that good bye Google of there assistant make calls to the combined with the wave vs so so where was we actually have G15 ordered that can distinguish between between the the white boy sandals computer tou in and Andy and the computer generated speech like they’re not including wavenet so we can tell difference between the human and the Year EA in a non-human Bible with the boiler technology with a meeting with a well thank you Google for calling but it’s

what is the next person doing it so yeah so what were the current limitations then because it sounds like it sounds like it is capable of detecting humans or recorded voice I saw your eyes or YouTube demo if you’re taking it works really well in noisy environment as well as so you actually authenticator device on a plane to pretty impressive but it doesn’t why doesn’t it work like what what are the limitations in and what do you think would be the year that the typical attack vectors for from you know from Bad Education different naturally imitation for all types of biometrics so you know a lot of Deacon labours cannot go through the through the fingerprint check because they actually do they don’t have anything about I leave it dark in the room speaker

official preview cognition right because if you’re trying to do the facial recognition it completely dark room then you will not be able to do it either t-shirt if you’re in the middle of the animal guns and Roses concert things standing right next to the speaker probably it would be difficult for you to go through the voice through through the metric system is also question the rationale of going to your bank at the application in middle switched it needs a penetration tester Z security firms that are paid to break into systems and identify vulnerabilities are these guys now I’m paying attention to voice interfaces and Consulting you guys on are you know what they should be doing I’m of a new field for your

security on what can you do with a what kind of testing that they can conduct out so what are the music you can use across asking you like that the people who are concerned about security but also the people who banks typically pay to break into the united trying break into their systems imagine that they would also be interested in me know now are you being able to handle this new boat modality but it’s something very different to that of typical come over security no security systems in place at the moment the fastest are you not relying on their skills in social e&e engineering and when you reply a water metre calculator don’t really care exactly what you’re saying but they just can’t confirm the disability kills daughter don’t work certificate rating is it the Bad Guys they’re looking for the easiest way to break the system

voice interface it just proves to be not the one where they should apply the resources I said it’s easier to just go after the human wherever you can make contact with the human than it is to try that hack your way through some technology samba true ok I’m about privacy then so you’re capturing people’s data it sounds like very very little data now I’ve seen as you said it’s only takes from 1 to 2 seconds to authenticate someone so bit on the pricey side doesn’t sound like you really need to capture that much data is that always the case and you and what do you do with the day to the you do capture the capabilities that will provide there they always up to you now so when will work with customers’ dinner with fashion students and someone it’s always in our consumers choice to do with this technology and witty and when you fix a choice do I want to use the voice biometric capability or do I need to remember my passwords and remember

my social security number and provide my address and the name of my answers dog that it’s much easier to go for you with my metrics in in Andover discussed earlier conversations experience was so so so so so people really warm this keep this simple easy operation with the Wii U Wii with integrated at Sunday you can only going to mechanic of the Year of the decision to the consumer makes so it’s always consumers decision to utilise this technology are probably going to go for the easier choice which is voice and they’re probably not gonna probably not going to question it too much and a lot of places like bounce probably insist on it or do it in the background without them knowing about it so this data is ending up in other people’s hands and how is it stored and does with gdpr and stuff to does a person always have control over it

can I decide to delete at any point and cannot how does that work in practice we as a company we we we never actually touch the consumer data so so when we deploy our technology of Wicklow it either on the on the mobile carrier application so if you’re using it on the mobile application that it means that you would would not actually leave your own device or if you’re using it as a part of the word quarter large corporate citizen get me the diary algorithms would be deployed on the corporate services actually not leave the four walls of that was awarded organisation that and sequencing of millions and millions on device you can do they get the edge

King parody Mrs the right term let you don’t require huge service to to process the data you are able to actually authenticate on device so that they can never leave you or your device is already within your within your position within your control ok well that’s wonderful

ok listening listen up if you’ve been listening to the show for a while now and feel like it’s time to contribute just go to voicetechpodcast.com / donate where you can contribute $1 a month $4 a month for $8 a month to help people show running This podcast wouldn’t be possible without generous contributions from lesson and such as yourself if you’ve never use patreon before we’re not take this opportunity to give it a go but made it super simple and secure to donate to independent content creators like myself and there’s never any commitment so you can cancel whenever you like you like to see the options my patreon page is that voicetechpodcast.com / donate patreon tiers are also available for businesses so if you have a voicetech product that you like to promote on the podcast newsletter and blog then you can become a voicetech champion at overturn voicetechpodcast.com / donate the more details ok so now let’s get back to this amazing episode

well another thing that would come to mine was that you know if I’ve if I’ve just signed up for my banks vital signs of medication as I’ve repeated that you know the password to the times I got my voice print or it’s sat on my device and then I go on to the next service and then asked me to the same thing and the same thing and it would be great if I could use use that one voice print everywhere right I’ve already authenticated it what you already know it’s mine is that on his is that available right now that on the road map on WhatsApp what’s the plan for that how’s it going to work because I’m right now also saw some devices do they have something better biometrics on them but we actually don’t see the Enterprise customers utilising definitely not all of them doing the reason for this is that

I show you talking out which for biometrics biometrics on on under which devices this functionality is already exists like fingerprint of stuff but a lot of people are not I’m not take advantage of it already added in the form in the order laptop the trouble with that is essentially you’re blind with trusting desiccation decision made by third party where’s the Oldbury liabilities state the definition or wherever it in the application so difficult for all them to agree to single standard organised single arm in a single solution service everybody’s so because he even servicing. Everybody none of the individual

the prices would have to control over and they and they do want to make sure that they can control it that they can monitor and the it’s on a support so so so so so so although there is a lot of simplicity in the user experience that you describing we would we still expect this market to be fairly forever fragmented for please for the next few years I say yeah that’s a shame really cos it seems like the technologies that are in place to to to provide that kind of service I was the one thing that comes to mind is is the blockchain you know if I kids control my voice print put it on the blockchain and end that was authenticated then you need that that could quickly I would imagine quickly become trusted in the in the way that have you no sign in with Google or Facebook is for the for the web I think in the end we will come up with some sort of a social like that but I don’t think that weird I don’t think that this is happening in the near future

ok 95.8 I then just just wait we can pretty touching that again you said it was really crucial to the both the authentication of the human or not and that the correct human what can you just give us a brief outline of washing machine learning methods you use in your company for both of those tasks and and and how they work in that some of the technical details at least we can understand how it how it works what’s a waterfall methodology were using what separates you not delete the Wheels on the spaced from from from the rest of the back so so so so so it comes to the

yeah I think they’re the number number of important components here so so so we will utilise both in our dinners and senans so deep relaxing convolutional networks depending on the particular task at hand there are a number of different architectures that are being deployed for 4 Hour 4.. I think one of the big am in a different raters for us is that we collect it and getting it would be all one of them probably one of the most unique datasets when it comes to the voice biometrics in the end and am anti spoofing so so we were able to train training algorithm the train and train our annual networks on the dataset Stata unique in in the Indigo

first city of the Beano types of discussion attack types of the devices that we used and therefore use or other recordings are types of the use cases so we really spend a lot of time in the afternoon on building this unique datasets stated that they can so so so literally so we can utilise them for training purposes what advice would you give to like developers and people looking to build this kind of technology into their into their voice apps today I mean first of all we can talk about how how to integrate ID R&D products into existing into a voice app for instance a skill or a mobile app and then if there’s any other advice you can give like some general, design or coding advice for the infant implementing this kind of the security software what are the most important things for the Communities to know that would it be for the primetime there Radio 4

for providing this security to the conversational interface in a completely transferred way it can be deployed in Aberdeen from ways we can provide our technology assists as the keys for Sunday, but not warm mobility is the casing and an Android in the ace of Base ATK stand linux and windows are we can also provided via the service and you and and and do it in developers seeking simply are linked through the apis and the appearance of we were happy to provide Saudi integration is eating someone is possibly not used to dealing with audio and what would be the advice for like preparing the audio as sending it to your servers and and also the considerations about the microphone like do you need a special microphone on the device of Fire far field microphone to be able to pick up the voice

from afar in noisy environments etc but we now but I don’t like the preprocessing so if you’re doing something for preprocessing for instance to enhance the ability of the signal or for the purpose of these are we actually would rather have the lost signal then the signal they went with some type of noise cancellation are so so so because typically cats out certain parts of the bandwidth that are not important from from alpha motability or the Acer perspective but actually quite useful for an hour so we talked about this in a previous Podcast episode as I know what we can’t hear we can you can compress of the sound right down but it they just rounded it uses for a machine learning algorithm and what about design

considerations on what is there are seen as a good way and a bad way to do authentication if this product works correctly and as we’ve described there is no that this isn’t it is nothing visible to the user and both in terms of developing it is that you have some kind of design guidelines of as to when when and where to capture the audio and the kind of things to prompt the user to say in order to get their the best result etc prediction of the what is there is a good diversity of the signal diversity special the valve that is the speakers using so I would like to have your number of samples of The Voice a good way to do it is to engage the user in a dialogue to collect

and then the way that our system is organised once we have enough data we we can provide that information to the interface and said well now we’ve have enough information to build the year are doing biometric template and from this point forward we can go ahead and use the app voice biometric also use a technique called immortal enrichment which means that with every single successful verification where should we recalculate the underlying biometric template and to make sure that we constantly in enhance the quality verification explain the Living sorry what’s it called that’s a modal enrichment ok so it goes through this occasion

would take the underlying biometric template in recalculated using the previous model and the new data that we received during the successive verification so so the more users are engaged in the system and Dr system the better quality artificial get right absolutely yeah and yeah I can again if you if this was that this was something that the user could possess and users across multiple services and then there in reaching their that the voice print that they have Latino to offer it to offer everybody so I can see that being a bit becoming a valuable asset for anyone who uses this thing becomes stronger over time alright I can love you listen to this thinking of this is huge map of Opportunity I’d love to get into the field voice biometrics and it’s not like we just said there’s so many so many possible applications are so many startups you could build just of the back

this one technology I think if I wanted to learn how to do this stuff what would be the best you know best best way to get into it would end what would be the the technologies that I should concentrate on first what is it more signal processing more machine learning there other other skills I need as well Hannah how would I do it if you really see the explosion of connective did the member companies and start-ups that I can entering the field I book a facial recognition there’s dozens and dozens of new companies that are they coming in with the weird new key capabilities of a pretty good quality but not seeing this invoice and the reason for his action of wishes is a much more difficult from a scientific perspective so so my advice will be to to to to study this even prophesying in the Beano speech and

definitely my alarm my finger in the end in Cin you know and then try to get into the field. But this is Anfield where the fact that you know we build a company round the Ivy Restaurant St E15 you know by PHP foreach JS without experience specifically within species that is a very strong competitor Spanish to us because we actually have the expertise to bring to the table and a really getting really shows in the quality of the technology that sounds like if you really want to be building this thing at the fundamental level you gonna need a team of people with that level of experience but in order to build a start-up on the back of some of these technologies mate there’s there’s many open source tools out there and there’s a lot of different ways you can get into it that don’t require how to get heels designer

connect to the apis and it’s a standard stuff but the indus indus in this one that’s one of the girls Adidas trefoil to make it easy no just put in users but also for the development of for the developers code every community that can include the security component into the voice of The Voice interfaces each of the voice I’m excited to see what people are going to be building with the Tangiers it starts to come to the fore what’s the Vision that keeps you excited about the field what inspires and inspires you to drive forward the ability to bring the same level of security the same level of comfort that we have when we talk to each other when we talk to her friends and family to bring the same

I keep ability to the children occasion PE28 asteroid between the users and the technology of being the customers in the year the Enterprise and it is really really is a new level bubble CCD security and convenience and being the first in the market Newbury with the social like that is really you’re really quite something so so so so so so we see a tremendous potential here excited to bring you new technology to the market there’s always new ideas that were bringing to the table Justin recently you know we are you know we launched a new capability that catalyses Diageo stream in order to identify key events so we can I do to find no dogs barking baby crying glass breaking in someone so so anyway that you guys

implementing like to straight into your into your product as well as just a future ok. Space but I think the promise of America right now just simplification of humans and they can I bring in some quickly and with security you no reply, tell me a podcast you’re single you’re tired of shopping anything especially if your shopping online you know it be a big a big part of it experiences you know constantly liking in your information answering me know security questions and soreness or for worse than making that experienced much easier to get your bring this issue to the conversation interface itself simplifying that experience so so I’m with you and what amor amor amor so tired of that my mum my traffic to do it through 303 normal conversation with a jump into hope so

language is excited to bring this capability I’m excited to I’m right behind you and I think I just know I can’t wait for the day where every microphone is is unavailable interface I can just walk up to I can I can talk to my friendly virtual assistant who knows me better than I do I think I can have a conversation with any Brand and buy whatever product or service that I like without have to jump through a lot of authentication Hoops and some of the some of the products I’ve been using my mobile app recently I can’t even get into them because of that sudden capture thing I’ve had to give up me you know if I let you know sending email support that is just ridiculous so I can’t wait for the day where you know we can just slightly laugh about it in a distant memory kids about you know the time we had to login now this is it ever going to believe anyway alright it’s been fantastic to talk to you Alexia good luck with everything and thanks very much for coming on the show and tell him it’s all about it

ok so you just heard from alexey keeper of the president at ID R&D at Salford today hope you enjoy listening you can find all the show notes with links the resources makes of the episodes at voicetechpodcast.com and that includes the conference links I’ve enjoyed this episode as many ways to support me in the show and you can tell one friend or colleague about the episode you can leave a quick review on iTunes that voicetechpodcast.com / iTunes and I promise to read those out and you can write for the blog. Forecast.com / published thank you to everyone who submitted their blog post I will get round to publishing those and insure older and you can also become a sponsor just like her hacks.com dash Barton her and all the rest and voicetechpodcast.com / donate are your contribution really does make the show possible so I’ll be back soon with another episode at until then I’ve been your house, Robinson at thank you for listening to the

Share this article

What do you think?

0 replies

Leave a Reply

Want to join the discussion?
Feel free to contribute!

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Related Posts

Marco Liuni Alta Voce
National Cancer Institute NFvdKIhxYlU Unsplash Julia Anderson
Christian Petroske ShoutWorks

Get notified about new articles

[yikes-mailchimp form=”2″]