Site:signifyd.com Why Machine Learning Is Not The Silver-bullet Solution To Online Fraud

There are better tools at the disposal of enterprises to combat online fraud than machine learning.

By Will Willie

Nobody predicted the big threat of fraud in online transactions. When you think about the scale of it, it is staggering. More than $200 billion of dollars is lost per year in card and non-card fraud. Most of this is happening on mobile, with fraudsters mostly stealing consumer banking credentials. These were the top three topics at your first AMA: New Machine Learning advances, Impact of Big Data on Fraud and How marketers should leverage Artificial Intelligence to help combat fraud

Has there been a shift in fraud efforts?

What happens when a hacker steals your data and turns it into money? How do you protect it from this inevitable theft? One of the issues we are seeing is that unlike banking fraud, fraud on mobile devices (by phone or tablet) is very different than fraud on physical fraud. With fraud for mobile devices is very different than on an ATM. Fraud on the digital domain is largely conducted online with mobile devices more than double as fraud for mobile devices as compared to last year.

Has there been an effort to address fraud on mobile?

We are seeing significant fraud enhancements from our technology partners. Fraud detection has traditionally focused on finding good actors (known good actors) and offloading the bad actors (unknown bad actors) to sophisticated systems and these are about halfway through the journey to the optimization. As the fraudsters are tested and have their services used, some new bad actors emerge in an effort to work around the optimization systems. We are starting to see implementations that are much faster and more accurate. Instead of being 50 percent accurate, our systems are more like 90 percent accurate. Again, with the good and bad actors, some bad actors will obviously emerge and their bad actors can actually be more effective than known bad actors so we are still behind in these efforts. Fraud is the manifestation of much larger problems and we’re just scratching the surface when it comes to working with fraud.

What big changes do you think we’ll see for fraud when hardware and software is a fraction of what it is today?

Most of the fraudsters are based outside the US. The way fraudsters were able to commit fraud the last time we really took big strides was in the 90s. In the late 90s, we had H.B. Fuller look at the hardware and software for anything that could be done. At that time they built a low-power calculation optimization engine to take our optical character recognition on PDAs and our voice technology and use it to take stuff like fraud. They had a very low-power version of it and it was just as effective as the original engines. Now, while we are going through the same kind of experience now, the stuff we need to solve is much more complex and we can’t just take an existing engine and do it. What we have to do now is say, okay, we have one problem to solve and we can take optical character recognition and our voice to perform this fraud and use that and split it into a number of pieces. This now provides good assurance.

Machine learning as a contributor to fraud enforcement is still new and you’re still pretty early on in the journey. What kind of fraud enforcement can we expect in future years when machines are capable of fighting fraud on a much deeper level?

I think that the “fraud of the future” is going to be caused by some combination of machine learning and human action. There is likely a lot of room for improved detection of fraud before it happens, but when it does happen, it’s going to be about the human.

No one expected the big threat of fraud in online transactions. When you think about the scale of it, it is staggering. More than $200 billion of dollars is lost per year in card and non-card fraud. Most of this is happening on mobile, with fraudsters mostly stealing consumer banking credentials. These were the top three topics at your first AMA: New Machine Learning advances, Impact of Big Data on Fraud and How marketers should leverage Artificial Intelligence to help combat fraud.

is a top SaaS/Real Time Marketing platform that offers a feature-rich e-commerce marketing solution, . Based in Houston, TX. Designed by marketing professionals for marketing professionals. They say it’s easy to use, there’s no license, and easy to scale.

Click to rate this post
Total: 0 Average: 0
  1. Home
  2. Career
  3. Site:signifyd.com Why Machine Learning Is Not The Silver-bullet Solution To Online Fraud

Related Articles

Leave a Reply

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

Fill out this field
Fill out this field
Please enter a valid email address.
You need to agree with the terms to proceed

Online Learning On How To Write And Format A Letter
“””where Does You Come From”” Learning English Online”
Menu