John Collison speaks to Siliconrepublic.com editor John Kennedy at the 2015 Web Summit in Dublin. Image: Connor McKenna
The Collison brothers’ $5bn payments powerhouse Stripe is making its move into machine learning in a bid to wipe out card fraud, which is expected to total $183bn by 2020.
Credit and debit card fraud has only accelerated as the internet has grown.
Employing machine-learning tools, Stripe’s new Radar technology is powered by the company’s own behaviour network.
This is a machine-learning system that learns from and adapts its defences based on transactions across hundreds of thousands of businesses, using the Stripe’s technology globally.
‘With Stripe Radar, businesses essentially have caller ID for incoming charges’
– JOHN COLLISON
According to Stripe, this is the first and only machine-learning-based fraud tool that requires no set-up and works for every user from the moment of their first transaction.
Thousands of signals collaborate to target fraud
Radar uses a technique known as machine learning, a type of artificial intelligence that allows computers to train themselves on data and apply learnings to new situations, without intervention or instructions from human programmers.
Relying on machine learning to analyse and weigh hundreds of different signals about each individual charge that Stripe sees, Radar uses advanced algorithms for detecting patterns and instances of likely fraud.
Radar grows and evolves as it is exposed to new data, meaning Stripe users’ defences against fraudulent transactions are continually adapting and getting stronger over time.
“With Stripe Radar, businesses essentially have caller ID for incoming charges,” said John Collison, president and co-founder of Stripe.
“Stripe’s behaviour network uses machine learning to learn from hundreds of thousands of businesses around the world running on Stripe, as well as signals from our financial partners.
“Because of this network, Stripe Radar can effectively spot patterns and detect fraud, protecting every Stripe user from the moment their first charge comes in.”
Stripe was started by Irish brothers John and Patrick Collison, and was recently valued at $5bn after raising close to $100m from investors, including card giant Visa.
Stripe is quickly becoming a part of the US and global payments infrastructure. In the past year alone, around half of all Americans have purchased something from a Stripe user.
Yesterday (19 October), we reported how both rival candidates in the US presidential race – Hillary Clinton and Donald Trump – between them spent $1.5m using Stripe services in August alone.
If a business started on Stripe tomorrow and received its first credit card payment, the chance that Stripe will have seen that particular card before is over 80pc.
As part of its intelligent fraud modelling, Stripe Radar will also take into account traditional fraud checks, like card verification code and address verification.
Within a two-month period during the beta, Radar was able to block more than $40m of attempted fraud for Watsi, a non-profit that helps fund medical treatments for people around the world.
