PDF link, for those who don't like videos: https://www.usenix.org/system/files/usenixsecurity23-oz.pdf https://www.usenix.org/system/files/usenixsecurity23-oz.p

PDF link, for those who don't like videos: https://www.usenix.org/system/files/usenixsecurity23-oz.pdf https://www.usenix.org/system/files/usenixsecurity23-oz.pdf If you were wondering about how they bypass browser permission dialogs, they don't: > the user needs to navigate to the malicious site and grant read/write access for browser-based ransomware to be effective Moreover, the browser's permission dialogs are pretty restrictive already - user can't choose things like "entire disk" or "entire My Documents", just a subfolder of those. So the paper focuses on things like encrypting Dropbox folder, which means user has to specifically select the folder, then grant access to that to the suspicious webapp. IMHO this is not something that I would be worried about, I don't think this will ever reach the scale of native ransomware. And I am certain that incorporating machine-learning based classifier into browser, which has "only" 1% chance of falsely blocking the app, is a horrible idea.