In an interview with TechRepublic, Joe Burton, CEO of digital identity verification company Telesign, discusses how the “fuzzy” area between statistical analysis and artificial intelligence is driving global, fast and accurate identity management. I was.
Telesign may have helped develop two-factor authentication, but it only has a small share of the market dominated by companies like Persona, OpenID, Okta, Duo Security and LastPass.
Burton said the company is looking to the future and has big plans to use new AI-powered technologies and services to differentiate itself from its competitors. A major approach since 2019 has been to evolve its communications platform-as-a-service leadership, do away with passwords, and focus on mobile numbers for identity verification, data modeling, and customized communications.
See also: 1Password’s Steve Won: Passwords coming soon Past tense.
Burton, who will become Telesign’s CEO in 2021, spoke about the company’s use of machine learning and how it provides security without adding friction to consumers.
TR: How important is analytics to identity management in terms of user behavior?
Burton: We see about 5 billion unique phone numbers flowing through our system every month in 195 countries on behalf of about 3,000 businesses, so we know what travelers look like. I have a very good idea as to whether a person’s identity has been stolen. We look at 2,200 different attributes of your phone usage patterns, all of which we use to train our blazingly fast and accurate machine learning models. A suite of explainable AI analytics can be used to respond to whether this looks like someone’s legitimate phone number.
TR: How important is the explainability part of this result?
Burton: I think this is going to be the foundation of AI for all materials in the future.Much the same as the [Advanced Placement] Tests in high school; if someone wrote down the answers and didn’t show their work, we don’t give them much credit.
look: For authentication over 5G networks secure network.
TR: Do your customers, your website or app, need to know where their decisions came from?
Barton: Back in the real world again. A good teacher may give a bad grade to a student who gives an incomplete answer. We should hold AI to the same standard.
TR: How do you generate analytics from your data model?
Burton: AI systems are built around the use of global, fast and accurate intelligence, and that person is who they say they are, creating an account on System X. It’s about how likely it is.
TR: Doesn’t it require you to have a lot of personal data on hand to do this?
Barton: We don’t have a huge corpus of data about a specific person. Instead, if you send a notification to a phone number, your location will probably change every time you see this phone number, [or] Perhaps the user is roaming on the Vodaphone network around the world. Feed it into an AI to create a new statistical model based on this phone number movement event in Europe. Then discard the data. But what we have is this interesting statistical model.
TR: So the data model you develop acts as a proxy for the user, rather than the actual data?
Barton: We don’t do database searches when we have new events. I have 20 questions. Let’s say this person tried to sign up for three women’s clothing sites in a row, which is atypical behavior. It’s not Valentine’s Day or Christmas. We send back a series of reason codes revealing that we saw a new practice involving a different type of new vendor (in a different location) than the person has done business with in the past.
TR: Why do we need AI for this kind of user view? Why can’t we do statistical analysis without AI?
Burton: There are several answers. First, there are many different AIs. For example, using logarithmic regression. This is an elaborate statistical analysis with a bit of AI “fuzziness” around the edges. I can say, “How statistically does this resemble normal behavior?” Is the activity around this phone number moving away from the normal statistical behavior of this user and similar to the activity of another cohort, say a bot farm?” You can’t get this from a purely statistical model unless you run Without AI, unless you’re really smart, you’d have to keep everyone’s personal and identifiable data. It’s just training a model, throwing data, training a model, throwing data.
TR: And how does this application of “fuzzy” AI improve the protection cycle?
Burton: It will change your life. If you hack us, I don’t want it, but you can get a bunch of statistical models. You don’t want me to keep track of everything you’ve done in the last 20 years. I don’t I have a stat number, a model, associated with a phone number, so when handed a new event, I use a series of reason codes to tell how I like or dislike typical behavior I can.
TR: What are the biggest challenges in authentication and digital identities?
Barton: Well, ID is an arms race. Please tell me your name. Please tell me your name and password. Now make your password longer and answer 3 secret questions, making it 7 secret questions. So it’s a mess. Telesign has a modeling idea based on this use case — is someone trying to create an account, sign in to an account, pay for digital goods on a gaming site, [or] Let’s hail a cab — being able to build global, fast and accurate models is like magic for us. Almost all mobile numbers in 192 countries use these models, so when you travel to Pakistan and try to log into the normal system, you will be left wondering whether this looks like you are in Pakistan or a hacker. You already have a strong concept. This is really important.
look: While the password still exists, avoid these are.
TR: What are you most excited about about Telesign’s progress?
Burton: We look forward to combining our Zero Trust stance with creating great customer experiences. A website makes you do 500 things to prove it’s you. I hate that idea. Frictional security is not the answer when moving into a digital context. Zero friction is also not the answer. The answer is to match the friction to your website. Creating the right amount of friction at the right time — that’s our mission, and frankly, it’s only through very clever use of AI that it really works.
