Pindrop, an Atlanta-based company that provides fraud detection software for phone calls, announced a set of updates to help its customers better protect themselves and people they serve from phone-based fraud.
One of the key updates is a new voiceprinting capability that identifies who’s on the phone and checks them against a set of people Pindrop has identified as fraudsters. The company is also providing a new Network capability to its customers that will provide real-time predictions about how likely a call is to be fraudulent, based in part on information gleaned from other Pindrop customers.
It’s based on Pindrop’s Phoneprinting capabilities, which will be upgraded with this release. The updates the company made to its models should be up to 20 percent more accurate than their predecessors. Phoneprinting works by analyzing call audio and using those factors to determine how likely the person on the other end of the phone is a fraudster.
Phone-based fraud is a massive problem around the world, since scammers can call from anywhere under false pretenses and try to get sensitive data, perform financial transactions and pull off other nefarious deeds. It’s an attack that relies on the human nature of the human on the other end of the call center phone — people who usually want to help.
Pindrop should help cut down on that by providing information like where a caller roughly is — so that it’s possible to determine if that doesn’t match up. (For example, it shouldn’t be possible for someone who’s in San Francisco to make a phone call from Ukraine.) What’s key to the use of machine learning is that it’s possible to identify where someone is based on other call characteristics, so that person imitating the correct caller ID shouldn’t fool the system.