Feature Spotlight: Piecing Together a User’s Identity
We love giving our users an opportunity to discover more about what’s going on behind the scenes. Sometimes we’ll hear one of our customers say, “Can machine learning really do that?” or “Wow, I didn’t realize that’s how Sift Science worked!”
In our Feature Spotlight series, we give you a peek under the hood of Sift Science and dig into the various features that make up our machine learning platform. In each post, we’ll examine one of the features that Sift Science analyzes to detect fraud – ranging from “Does the user share a Device Fingerprint with a known fraudster?” to “transaction velocity over the past day”. Check it out and send us your questions!
In the world of fraud detection, one of the most important checks that you’ll undertake is verifying someone’s identity. “Is this a real person?” “Can I trust that they’re actually who they say they are?” And at Sift Science, we’re invested in helping you answer these complex, crucial questions as quickly and confidently as possible.
For one thing, we aim to make it easy to verify a user’s identity using our web console. On the User Details page, there’s an entire section specifically dedicated to showcasing identity-related information so you and your team can authenticate a user. What names are listed on their account? How about phone numbers? Do they have any presence on social media?
But that’s just the tip of the iceberg. Behind the scenes, our machine learning technology is analyzing a bunch of identity-specific signals and incorporating those learnings into our predictions.
To start, as it’s sifting through tetrabytes of data, our platform automatically detects whether each particular piece of that data is related to a user’s identity and how, so our machine learning can make the most of that information.
Here are just a few examples of how Sift Science’s machine learning uses this info:
Linking identities
Has a single user created multiple accounts on your site? Our machine learning technology can link together identifying information like name, email address, and social media information – in addition to behavioral and device-specific signals – to let you know they’re actually the same person. We can even match variations on names, like Bill, William, Will, Billy, Willie…you get the picture.
You may have seen our recent blog post covering typos and word choices that are more likely among messages posted by fraudulent users. Well, we can also use these postings and messages to help link accounts together, based on similarities between the wording they use.
Name changes
Billing address name keeps changing? That’s a major red flag for credit card testing on an account. We track all name changes (as well as other key signals like updated credit card information), and they weigh heavily in how we determine whether a user’s up to no good.
Phone numbers
Similar to how our machine learning technology analyzes email addresses, phone numbers offer a goldmine of useful information for detecting fraud. We extract as many signals as possible, such as whether it’s a landline or a mobile phone, the number’s carrier or network, and geographical information based the area code.
Identity signals are indeed crucial for detecting fraud, but they’re just a few of the thousands of signals analyzed by Sift Science’s machine learning platform. Want to learn more? Check out some of the other posts in this series, like how we analyze user behavior patterns to help detect fraud.
