Google Chrome has built-in phishing detection that scans pages to see if they match known fake or malicious websites. Now that the technology has been hardened again, Google is announcing that in Chrome 102, running machine learning locally (without sending data back to Google or elsewhere) can help identify sites that are actively asking for permission to notify, and send notifications before they pop up. it’s mute.
For this technology upgrade, Google explained
To further improve the browsing experience, we’re also improving the way people interact with web notifications. On the one hand, page notifications help deliver updates from sites you care about; on the other hand, notification permission prompts can be annoying.
To help people browse the web with minimal disruption, Chrome predicts when permission prompts are unlikely to be granted and silences those prompts. In the next version of Chrome, we’ll be rolling out an ML model that makes these predictions entirely on-device.
In a future version, Google plans to use the same technology to adjust Chrome toolbars in real-time, showing different buttons like the share icon or voice search when and where you might be using them, without adding extra tracking calls home to Google or anyone else. If you prefer to manually select buttons, that’s fine too.
