Google Chrome has built-in phishing detection that scans pages to see if they match known fake or malicious sites (using more than just the URL, since scammers rotate those more quickly than it can keep up, and now that tech is getting better. Now Google also says that in Chrome 102, it will use machine learning that runs entirely within the browser (without sending data back to Google or elsewhere) to help identify websites that make unsolicited permission requests for notifications and silence them before they pop up.
As Google explains it, “To further improve the browsing experience, we’re also evolving how 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 become a nuisance. To help people browse the web with minimal interruption, Chrome predicts when permission prompts are unlikely to be granted, and silences these prompts. In the next release of Chrome, we’re launching an ML model that is making these predictions entirely on-device.”
In a future version, Google plans to use the same tech to adjust the Chrome toolbar in real-time, surfacing different buttons like the share icons or voice search at times and places where you are likely to use them, without adding additional tracking that phones home. And if you prefer to choose your buttons manually, that’s still going to work too.
Last summer, Google announced performance improvements as part of the Chrome M92 update bringing the time to calculate phishing classification results down to 100 milliseconds from 1.8 seconds. A blog post today reports that in March, it updated the machine learning model to identify 2.5 times more sites that could be phishing attacks or that try to deliver malicious downloads. Google also says it updated the browser-based machine learning tech behind its Journeys feature to take you on a trip through old searches, so that now it will figure out if the language of the page needs to be translated.