What would we do without notifications? These little messages are part of our daily routine: they can come from apps, of course, but also from websites. Sometimes useful, sometimes invasive, and often annoying, these alerts can also hide a much darker side. The most shady websites can thus push Internet users to subscribe to their notifications, which gives them an additional means of scamming them.
Booty Notifications
Google has foreseen this. On Android, Chrome has launched an alert system against notifications deemed misleading or unwanted. With the help of Gemini Nano, a locally running AI model, the browser can now identify suspicious notifications—such as those that encourage downloading dubious apps or sharing personal information—and warn the user before displaying them.
When a notification is deemed problematic, the user sees a warning message with the name of the sending site and can choose to read it, unsubscribe from it, or permanently allow notifications without future warnings.
This detection system, designed to preserve privacy (no data is sent to Google), was trained using synthetic data generated by the Gemini model and validated by human experts. For now, the feature is exclusive to Android, but a wider deployment is planned.
At the same time, Google published a short report on combating online search scams. We learn that hundreds of millions of fraudulent results are blocked every day by Google's systems, thanks to updates rolled out throughout 2024. With its AI models, the company claims to have increased the number of suspicious pages blocked before they appear in search results by a factor of 20.
Fraudulent queries related to airline customer service are down 80%, after the implementation of specific protections. Fraud attempts imitating official sites (for government services, for example) are down 70%.
Source: Google
0 Comments