http://blogs.msdn.com/b/oldnewthing/archive/2014/12/17/10581...
It will be news if there's an account recovery mechanism that can be exploited by a total stranger without access to your copy of Chrome/FF.
You would expect that 2FA could be bypassed for the locally cached password database, since AFAIK there is no such thing as requiring an OTP to decrypt the ciphertext that you have, only to prove to a server that you know the TOTP secret.
Lastpass discloses that you can turn off 2FA for new logins by proving ownership of an email address. You can set a separate email address for this purpose, and I believe also turn it off.
It occurs to me as I write this that for most people compromising an email account would be just as disastrous, given how many services rely on a simple email for password recovery/resets.
I think the difference between someone who uses a password-manager user, and the vast majority of people who don't, is that those who don't have HUNDREDS of points of failure that contain the keys to the kingdom, some completely insecure.
Not that it doesn't have its own vulnerabilities.[1] But perhaps those are more bugs (e.g. not enforcing sandboxes properly) than architectural weaknesses.
[1] http://arstechnica.com/security/2015/06/serious-os-x-and-ios...
Keep your important passwords encrypted with GPG on flash card, with backup on an another flash card. Regular GPG works flawlessly in terminal. Small shell script, which will ask for password and site name to display site password(s), will do the job on any Linux-compatible smartphone.
Instead, I just use KeePass, which I protect with an offline SSH key as well as a strong password. In order to compromise my passwords, you'd need my KeePass database, access to one of my devices (laptop, desktop, but not phone), and my password.
However, this means you can potentially recover vaults from unattended / stolen / lost / unwiped hard drives, phones etc.
The authors talk a lot about how you can get everything if you compromise the client machine, but of course if you compromise the client you can compromise everything.
Just install a keylogger to get all typed passwords, send the decrypted hard drive image over the network and run it in a VM proxying network traffic via the compromised machine, no need for any targeted attack.
The fact that they focus on that rather than on an actual interesting attack makes me believe that there is no real vulnerability whatsoever.
"We also found how it is possible to abuse account recovery to ultimately obtain the encryption key for the vault" is really scary, but if it were accurate they would be talking solely about that, so maybe it's just a phishing attack.