Robots flip burgers, make cocktails, engage in the practice of law and perform an increasing number of tasks all across the global economy. What if they were able to better secure consumers as well as their online payment and ID data — which, in turn, would protect companies from the reputational and financial costs that follow breaches and hacks?
That’s part of the general thrust of an upcoming PYMNTS webinar with Karen Webster and Sunil Madhu, founder of identity verification and fraud prevention services provider Socure. The webinar, scheduled for 1:00 p.m. EDT on Thursday (July 19), will include discussion about how a robust AI robot could leverage the full power of machine learning paired with massive data sets, drawing in data from online, offline and social sources.
That’s the motivation Socure had in mind as it endeavored to build its Identity Verification Robot.
Crafting Rules
To understand what AI and robots can do with online security, one must first understand that it’s all about the rules. Manual fraud prevention and ID verifications systems depend on rules crafted and executed by humans, which, as one can imagine, leaves all types of opportunities for criminals and others bent on exploiting loopholes. Those rules reflect a fair amount of human intuition, which is impossible to divorce from the rule-making process.
Older machine-learning methods, relatively dependent on human input, also cannot escape the limitations of human intuition. For instance, a rule to flag a source that opens, say, more than 10 accounts a day can be defeated by smart criminals who create one fraudulent account every 10 days.
“Eliminating bias is still an interesting problem,” Madhu said. “You can train a computer to eliminate those biases much faster than you can train a human being to eliminate those biases.”
As will be discussed in the Thursday webinar, robots can, in the words of Madhu, “do things more efficiently” when backed by cutting-edge AI and machine learning technology. That’s because the robots built with that technology can define and write their own rules, going beyond intuition to spot patterns that might be too subtle for a person to see — or least see in time to prevent fraud.
Robot Limits
Granted, there are limitations. Robots are not sentient, and that’s why they cannot discern cause and effect, Madhu told Webster. Granted, human beings are often terrible at that exercise, but the machines can help “identify correlations,” he said.
He used an extreme example to demonstrate the point: A robot can figure out, without too much digital sweat, a “stupid” correlation that might attempt to tie the number of people buying Teslas with the moon landings.
Such abilities help robots fight fraud and identity thieves, finding patterns that indicate criminals activity. And analysis of those patterns can, in fact, help with the cause-and-effect problem, assuming humans are involved. “We can study the effect and identify the cause,” which could be a data breach, Madhu told Webster during the interview, in advance of the Thursday webinar.
Multiple Fronts
More specifically, the webinar will focus not only on those topics, but on the current state of digital identity verification tools, and how AI and machine learning approaches comply with stringent regulatory requirements.
Fighting fraud requires a sustained approach that takes in multiple areas — PYMNTS, in fact, recently did a deep dive on call center fraud prevention.
No matter the approach that criminals and identity thieves takes, the cost to victims can be enormous. Recently, a federal judge ordered PricewaterhouseCoopers (PwC) to pay $625.3 million after failing to detect fraud between a client and a mortgage lender. The Thursday webinar will offer fresh ideas and information about some of the most up-to-date methods of blocking those criminals.
Enter your work email below to join Socure Founder Sunil Madhu and PYMNTS CEO Karen Webster on July 19, 2018 at 1:00 PM (EDT) to learn about:
We’re always on the lookout for opportunities to partner with innovators and disruptors.
Learn More