Microsoft has used machine learning to help the US Federal Trade Commission (FTC) crack down on tech support scams. A tech support scam typically starts with a pop-up that alerts users of a malware infection on their computers. This pop-up is fake, and prompts the users to call a tech support number supposedly belonging to a well known technology company. Scammers posing as telemarketers then charge the gullible users for expensive and unnecessary computer services.
A majority of the users targeted are above the age of 50, but 30 percent of the users were under the age of 49. Microsoft's Digital Crimes Unit receives more than 10,000 complaints every month by those affected by tech support scams. Experts believe that only a fraction of the tech support scams are reported by the users. The end users are not experts, and are not likely to capture screenshots of the original pop up. Additionally, each end user only has very little information to provide to the authorities, such as a number used for the scam.
The perpetrators of the tech support scams move fast. They keep changing the IP addresses that they work out of to make tracking them harder. The phone numbers used by the fake telemarketers are also discontinued after some time. Microsoft scoured the web for content that matched the approaches used by the tech support scammers. It would have taken a long time for humans to piece together all the information, and find meaningful patterns that could lead to the masterminds behind the tech support scams.
Microsoft got in the help of machine learning to do track the scams. The content from questionable sites were scanned with Microsoft's computer vision API and Microsoft Cognitive Services. Chris White, a principal researcher at Microsoft’s Redmond, Washington, research lab says "What we’re able to do is address the problem at the scale it’s happening, and provide the mechanisms for us to do something about it."
After tracking the patterns in the tech support scams, the researchers also used a data visualisation software by Microsoft called Power BI to present the findings of the AI scans to the law enforcement authorities. Doing this allowed the authorities to see how old the targets were, in which geographic areas they were located in, and what tactics were being used by the attackers. The findings allowed the FTC understand exactly how tech support scams were affecting people, and helped them in their crackdown.