Researchers Reveal Method To Stifle Malicious Robocalls
A method called SnorCall has been developed, utilizing artificial intelligence to analyze the content of robocalls. This AI system assists law enforcement and other relevant parties in identifying the sources of these malicious calls. Researchers presented their findings at the Usenix Security Symposium in Boston.
During their study, the researchers collected 232,723 robocalls spanning over 23 months, using more than 60,000 phone lines provided by Bandwidth, a telecommunications platform provider. These robocalls were transcribed and examined using a machine-learning network known as Snorkel, which gives SnorCall its name.
Snorkel enables the researchers to create and manage training datasets without requiring manual labeling, significantly reducing the time and effort needed for analysis.
Sathvik Prasad, the first author of the paper and a Ph.D. student at NC State, explained that Snorkel serves as the foundational layer for analyzing the audio content of robocalls.
The system is adaptable enough to categorize various types of robocalls. The researchers demonstrated this by focusing on harmful calls like social security and tech support scams, which have malicious intent.
While there are also bothersome but less harmful telemarketing calls such as auto warranty offers, the primary focus of the study was on the most harmful types of robocalls, like those impersonating border patrol or social security authorities.
The researchers aimed to understand and dissect the most egregious calls that pose the greatest harm to society.