i’m honestly not to sure how one would go about it, i know one of my friends has done it. I would assume there is at least one open source project for this type of thing. Realistically i can’t imagine it would be that hard, there are probably writeups on people doing it already. In the most simplistic form you’re keeping spicy private keys on an encrypted flash drive. That way they’re a physical hardware item, but also physically isolated. Though you would absolutely be in a bit of a bind if you ever lost it. Realistically, changing the key and it’s encryption will solve that problem though.

I’ve recently thought of doing similar things using forward secrecy keys stored on the flashdrive itself so that way it’s always different. Similar immediate security risk there, but again changing the key is the solution. Theoretically you could also do a two part key system, where you store a portion of it on your system, and the rest on the drive, so that way in the event of compromise, they only have a portion of the key. And they still need the other part in order to do anything.

scripting wise, it should be pretty simple, you plug in the drive, automount it, rip the key out, stuff it to where it needs to go, and then remove the drive. Always make sure you have secondary backups though, whether written down or stored somewhere. Losing accounts is no fun.

I’m not a security researcher or expert though, there are definitely smarter people out there that have already talked about this kind of thing at length.

What I did was to use keepass to store most of TOTP and use Yubikey to unlock it. Absolute critical ones like email is saved directly in Yubikey.

I use an OnlyKey and Mooltipass interchangeably. Prefer the lower tech OnlyKey. My passwords are half memorized passphrase and half random characters on the device. Only use for disk encryption, main account, and password manager.

So I get very confused over which protocol is which. I think the cheaper keys lack support for OAUTH. Which is required for things like windows login.

I suggest having a threat model about what attack(s) your security is protecting against.

I’d suggest this probably isn’t giving much extra security over a long unique password for your password manager:

• A remote attacker who doesn’t control your machine, but is trying to phish you will succeed the same - dependent on your practices and password manager to prevent copying text.
• A remote attacker who does control your machine will also not be affected. Once your password database in the password manager is decrypted, they can take the whole thing, whether or not you used a password or hardware key to decrypt it. The only difference is maybe they need slightly more technical skill than copying the file + using a keylogger - but the biggest threats probably automate this anyway and there is no material difference.
• A local attacker who makes a single entry to steal your hardware, and then tries to extract data from it, is either advantaged by having a hardware key (if they can steal it, and you don’t also use a password), or is in a neutral position (can’t crack the locked password safe protected by password, don’t have the hardware key / can’t bypass its physical security). It might be an advantage if you can physically protect your hardware key (e.g. take it with you, and your threat model is people who take the database while you are away from it), if you can’t remember a sufficiently unique passphrase.
• A local attacker who can make a surreptitious entry, and then come back later for the results is in basically the same position as a remote attacker who does control your machine after the first visit.

That said, it might be able to give you more convenience at the expense of slightly less security - particularly if your threat model is entirely around remote attackers - on the convenience/security trade-off. You would touch a button to decrypt instead of entering a long passphrase.

Oh that’s awesome! Thanks for letting me know

External entry of the pin, means you avoid compromising it on a compromised computer.

Me too on the „need backup“. Any idea how to go about that? I know some sites have backup keys for otp but I have no process for storing then and avoiding a bind (like storing the 2FA for my vault in my vault and getting locked out).

I will probably have to play through scenarios or is there a comprehensive guide on this (probably)?

I have one and I would not consider myself paranoid. I go to school part time and have to login with different accounts on rotating computers. It is nice to have a password manager I can plug into the PC instead of typing it off of my phone or having to memorize it.

Manually keying in the pin is only needed when plugging in the device. Challenges for TOTP, FIDO2, etc. are a configuration option, and are only 3 digits if enabled (press any button if disabled).

As for “excessive amount of security”, security as an absolute measure isn’t a great way to think about it. Use case and threat model are more apt.

For use case, I’ll point out it’s also a PGP and SSH device, where there is no third party server applying the first factor (something you know) and needs to apply both factors on device.

For threat model, I’ll give the example of an activist who is arrested. If their e-mail provider is in the country, they can compel the provider to give them access, allowing them to reset passwords on other more secure services hosted outside the country. The police now have the second factor (something you have), but can’t use it because it’s locked.

I think “unnecessarily over-the-top” is a key demographic in every market. Not a large one, but definitely present.