Joseph But it doesn't fix it. Even if offline, if the NSA has put a pattern in to the CNG of Windows for example and you use the randomness from that machine, they may be able to defeat it and thus defeat your one time pad. Granted, you might be able to get around this by having multiple systems work together to form the one time pad, all offline.

Joseph No, if the computer you write the message on isn't attached to the internet, you don't really care that much if it's a bad actor as long as it isn't expected that it has updates that might specifically target your scheme, but if you are going to question the security of the hardware in general, you should also be questioning the security of whatever you are using to generate your pad. If you do accept things at the algorithmic level but not the client/protocol level, then you can get pretty close to the same by just nesting encryptions of a root key and use something like AES as well. Mostly I was just pointing out that if you don't trust the cryptographic principles of your own clients, you still have a single point of failure on the cryptographic principles of the offline "magic wand", if you do trust the cryptographic principles of the machine, then you don't need the in-between steps necessarily.

R. M. yeah, to be clear, I think your devices are a good idea. My point was mostly that this original post isn't really that useful of an idea. Much better to harden the password storage instead, which is exactly what an offline device does for you.

For that system, you have to have a system you trust to make a cryptographically secure one time pad. If that software or hardware is compromised your entire system falls down still.

It really isn't good advice though. Having offline generators and rings is a good idea. I suggested using 2FA and a Yubikey for exactly the same reason. Simply adding a bit of complexity to the "what you know" factor doesn't add much security because the means of compromise of one mean the other is also likely compromised, as well as some other problems I mentioned in my own main post.

Phani Karan Well it's also not adding a factor. 2 passwords is not 2FA. It has to be 2 different factors. A master password for a password manager and a mini-password are still just different version of "what you know" and are just a slightly more complicated single factor.

Except that salting is only intended to make the use of rainbow tables inefficient. It doesn't add a security benefit beyond making it harder to attack an individual password. In this case, it's more accurate to say that in the compromised password DB scenario, this provides a very weak password as a protection against the main password being converted to effectively being a salt. Unfortunately, that's not likely to do much if someone has both a compromised hash table and your main password and the simple "horcrux" would rapidly be found by brute force hashing given it is short and rememberable.

I'm not seeing what this adds. This is the same thing as putting up 2 passwords and calling it 2FA (which it isn't). Under the vast majority of threat cases I can think of, you are either still screwed because they capture your login at the same time they capture your master password or you are stuck back with the problem of remembering a secure, non-guessable element for every account, which is exactly why password managers exist in the first place. This is either ineffective, impractical or both depending on the threat you are trying to protect against. It's also trying to solve a problem that was already solved. This is exactly the reasoning behind 2FA/MFA and making sure that access to multiple things is needed. It would be much better to get a Yubikey or similar and put your HOTP/TOTP secrets in to that so that it is physically isolated from your phone and the internet and thus can't be compromised without direct theft of multiple devices. You get the same advantages without the disadvantages and limitations of hanging your security on two of the same factor.