Pindrop – Voice Authentication and Voice Verification
Voice authentication and voice verification methods may have some obvious use cases – when you call your bank for example – but there is a far more important need for such solutions. That’s because there’s an interesting scam going around these days which targets the elderly. Here’s how it works.
First, the scammer looks for some tool out there who thinks it’s a good idea to share their life story to the public on social media. Let’s call this individual “Johnny.” Using social media, the scammer figures out Johnny’s dad is named “Pat” and Pat’s mom (Johnny’s grandma) is named “Sue” who happens to be in her mid-70s and lives alone. The scammer also sees that Johnny just sent out some pics in the past hour from a full moon party in Thailand, so the scammer decides to call Grandma Sue and the call goes something like this:
- SCAMMER – Hi, Grandma. It’s Johnny.
- GRANDMA SUE – Johnny who? Who is this?
- SCAMMER – Grandma, it’s Johnny, your grandson. You know grandma, Pat’s son. I’m in real trouble and I need help. I’m in Thailand and in trouble with the police.
- GRANDMA SUE – Ah yes, Pat did say you were going to Thailand. That’s horrible! Well, why don’t you call Pat?
- SCAMMER – I can’t Grandma. I don’t want my Dad to get angry because I got in trouble for smoking some marijuana. I just need you to send me $1,000 to pay the police and they won’t throw me in jail. I’ll pay you back. I promise. Please help me, Grandma! They’re all so corrupt and I need cash fast!
It’s at this point in time that a persuasive individual just might be able to convince Grandma to send over that $1,000. If this scam works one out of 10 times, the success rate is high enough to make the effort worthwhile for professional scammers – like the Sakawa Boys. The reason we know about this scam is because one of our writers actually observed it in person.
To make matters worse, there are companies we’ve come across before like Lyrebird that can take a 30-second digital recording of you speaking and then use it to mimic your very own voice. Just 30 seconds of your speech is all that’s needed for scammer Johnny to sound just like real Johnny. How many people get on YouTube and blab on about their uninteresting lives for hours, giving these scammers even more fodder for these sophisticated scams? Fortunately, a company called Pindrop may have the technology needed to put an end to these very sophisticated scamming techniques.
Pindrop and Voice Authentication
Founded in 2011, Atlanta startup Pindrop has taken in nearly $213 million in funding from a whole slew of investors including large corporate investors like Goldman Sachs, Google, and Citi. The latest funding round, a $90 million Series D, just closed days ago, and will be used to further develop their technology platform that promises to revolutionize how voices can be analyzed in call centers around the globe to increase efficiencies and prevent fraud. It all started back in 2011 when the founder, Vijay Balasubramaniyan, tried to order a suit in India and his bank flagged the transaction as fraudulent – then called him and wasn’t able to authenticate him. (Tenured travelers will immediately relate to just how pissed off this poor man must have been at that moment.) Let’s talk about retail banking call centers and traditional authentication methods.
Knowledge-Based Authentication Sucks
We’ve all been through those dreadful “Knowledge-Based Authentication (KBA) methods” which ask you questions that you’re probably unlikely to remember how you answered. Odds are, most people have more than one “favorite childhood friend,” more than one “favorite restaurant” that might (gasp) change over time, and more than one “favorite professor who changed their life.” We’ve all tried to struggle through remembering how we originally answered these questions several years back, and an estimated 10% of all people fail to answer these questions correctly. They also don’t serve their purpose, since an estimated “60% of fraudsters can answer KBAs at an estimated success rate of 60%, thanks to the rise of data breaches and social engineering.” This is just one of the reasons why call center fraud is on the rise:
Then there are the multiple authentications that take place when you’re handed off from John-in-Mumbai to his supervisor in a developed market, and they have the gall to ask you the same damn questions all over again when all you want is some decent customer service in an era where we’re creating synthetic life. That’s where “passive voice authentication” comes into play. Why not authenticate customers based on their “voice fingerprint” which can be determined the next time they call in, so that passive authentication can be performed for subsequent calls? That’s what one of the largest retail banks in the U.S. decided to do when they adopted the “Pindrop Passport,” a passive authentication technique which allowed them to save 55 seconds of time for each call resulting in $2.75 million of annual operational savings. With solutions like Pindrop Passport, it should only be a matter of time before KBAs go the way of pagers and payphones. While this is an obvious application of voice verification technology, there are other uses as well.
Voice Verification and Fraud Detection
Some of the more sophisticated solutions that Pindrop offers really go beyond what’s intuitively possible. Of course, the idea of passively authenticating someone using their voice is pretty easy to understand. But what about analyzing 1,300 different data points during a call to uncover things like where a caller is calling from, or using the method in which keys are pressed to determine if a bot is calling? It’s called “Phoneprinting”, and it’s really remarkable because it probes lots of data points that are impossible for the caller to control:
As it turns out, the fraudsters out there are using their own Robotic Process Automation (RPA) techniques to plow their way through automated systems using “automated bots to test large numbers of stolen credentials and credit card numbers.” This is especially problematic when you consider that in the future, “Gartner predicts that customers will manage 85 percent of business relationships without humans by 2020.” The extent to which scammers are getting more sophisticated is evident in the increased volume of scams conducted over time:
As you can see, the problem spans multiple industries with some more subject to problems than others. Not only that, but Pindrop lets you cross-check a caller’s voice against Pindrop’s “voice blacklist” to detect repeat fraudsters. These advanced solutions are one step ahead of where the scammers are which is where you need to be as an organization these days.
Pindrop takes 650 million calls a year, then analyze 1,300 different data points on each call. That’s a ton of “big data” that Pindrop can then feed to machine learning algorithms to obtain a competitive advantage over the other AI startups out there that are probably thinking about doing similar sorts of things for voice authentication and verification. Couple that with having corporate investors with deep pockets and their own call centers that could use such a technology, and you have what seems like a winning story. Eventually, such a solution could be brought into people’s homes as suggested by Pindrop’s partnership with Amazon. Maybe the next time “Johnny in Ghana” calls Grandma looking for cash, we’ll have the tools we need to screen such calls so they never reach her in the first place.
While NLP is now mainstream, there's plenty of growth left to be had. Four of the five AI stocks we're holding are SaaS businesses selling AI-powered enterprise software. Become a Nanalyze Premium annual member and get immediate access to our entire tech stock portfolio.