Hacker News Reader

Comment by tripletao | Hacker News Reader

What's the benefit of this? You have to trust the website to faithfully scan the dice and not send their state anywhere else. If you trust the website, then you could just wiggle the mouse for entropy to seed a cryptographically-secure PRNG, generate a password in the usual way, and write the password down (or memorize it, or print it, or print it as a QR code if the scanning is important, etc.). That seems equally secure, and more convenient than carrying a set of dice with me everywhere I go.

And of course a YubiKey or equivalent is the real answer. I'm trusting the developer there too, but I get a nice compact package that fits on my keyring, and a secret that never goes outside that package.

View on HN

> wiggle the mouse for entropy

You don't have contact info in your profile, would you mind shooting me an email at mouserng-hnpublic@lucgommans.nl? As far as I could find, there has been no research into the randomness of those key-generating mouse movements, so I'm slowly (over months, probably will be years in the end) collecting data to finally do this. Collecting a trace for me shouldn't take more than two minutes of your time. (Anyone else who's interested is also welcome to mail me, by the way. The address will work for at least the next 7 days.)

I agree that's an interesting and under-explored question. Perhaps you could gamify it, build a website where people compete to generate the maximum bits of entropy in ten seconds, as measured after compression by an algorithm that you continuously retrain and enhance based on that user data?

Yeah the plan is to eventually make it public, but for the moment I want to have a reasonable amount of data to make sure I'm capturing the right things. I mean, (x,y,time) is all there is to it, but perhaps I should be measuring it in some desktop software (where I might be able to capture the interrupts and their jitter) rather than just record the movements in javascript.

This is not a replacement for YubiKey, this is a system to back them up. The article says: "You can then use that key [...] as the seed to create a U2F key for two-factor authentication"

Imagine that you do not trust electronics. Maybe once in the past, your PC's power supply blew up and fried every device in every USB port. Or maybe you were by the ocean once, and salt water got into your pocket and ruined every electronic device there. Or maybe you have read too much science fiction and are super afraid of terrorists setting of EMP bombs.

Either way, you want a key backup which is non-electronic in nature. What are you options then? Not many. You can generate a random QR code and and print it. Or you can write down list of 13 random words, bitcoin-style. But this is still paper, so it can get lost or easily destroyed. The box of dice, on the other hand, is much bigger (so it is harder to lose), and much stronger physically -- it does not care about water damage or mold.

Yes, this is not for everyone. Some people will want to backup using another electronic device, or maybe they do not care about backups at all. I'd personally use the "13 words" method if I had this problem. But I can easily imagine a person for whom the dicekeys would be the best secret backup solution.

Maybe, though that seems pretty fringe to me. I'd note that backing up the secret from a challenge/response token creates a big new class of attacks--it's much easier for someone to learn my secret without me realizing when that secret exists outside tamper-resistant memory. It would be better to register multiple different security tokens, or to program multiple security tokens identically and then destroy the key material afterwards (except for the copies inside the tokens).

> You have to trust the website to faithfully scan the dice and not send their state anywhere else.

no, you don't. a) it's open source b) it runs locally

so, yes, you have to trust the software to some degree but at least you can audit it and it _should_ be relatively small amount of code. Also, you don't need to trust any remote site or the network because it doesn't use those and that's easy to test without looking at any code.

You can't audit most password managers and even if you could it would be a pain in the butt because they're so much more complicated.

> And of course a YubiKey or equivalent is the real answer.

as others have pointed out this is used to seed the FIDO2 key which makes it _better_ than yubikey because you can't really replace a lost yubikey.

They promise open source in future, but for now it's just megabytes of JavaScript delivered by their website, much of it generated by emscripten and not human-readable. In any case, they're encouraging people to use the website now. Given that encouragement, I'd expect most people to use the dice through the website, even if/when a standalone app becomes available. Even if/when you had the source, this would be orders of magnitude more effort to audit than a simple PRNG-based password generator.

KeePass is actually open-source now, and has been audited professionally. You are correct that a full password manager is more complicated than the dice; but the dice don't replace your password manager, so it's an addition to your attack surface, not a replacement.

And no, storing your secret outside the the tamper-resistant memory of a security token isn't better than registering a second (or third, or fourth, or however many makes you comfortable) backup token. HSMs and security tokens exist for good reason, and a cleartext backup loses that benefit.

I don't see anything wrong with marketing this as a cool toy, and as a tool to teach people how computer security works. I'm disturbed to see this marketed as if it solves a real problem, since that's at best a distraction from the steps (e.g., using a password manager and a hardware security token) that will help people practically.

Ok, but still, since you’re trusting the app why not have it generate entropy by mouse movement? I imagine that’d be just as unguessable.

Ok, so how do you backup your mouse movements?

By memorizing it as a passphrase, or writing it down, or printing it as a QR code, or saving it to a flash drive, or (best) saving it in the tamper-resistant memory of a security token protected by a PIN or biometric? Unprompted, I don't believe many people asked for the most practical way to securely store 192 bits would answer "the orientations of twenty-five cubical dice in a plastic box".

I like the physical nature of it. Cryptography is so abstract. I like idea of master password generating code device being a physical object.

The physical route is a good idea, but if you're going to do it there's no reason to introduce a weak link in the chain of having an app scan the dice. You can just get a set of 5 dice and do the lookups yourself using a word list like this [0].

[0] https://www.eff.org/deeplinks/2016/07/new-wordlists-random-p...

The author of that post is on the DiceKeys advisory board.

At some point, you're going to feed the entropy source into software. Having the app be part of your trusted computing base reduces the chance of a data-entry error that could leave users forever without their data.

Feeding the entropy source into software might take many different forms depending what is the password to (e.g. login/full disk encryption key on your computer, password manager master password, password to a specific online service, etc). For most uses you'll have to enter it as text, and have to enter it multiple times, potentially on different devices that you are accessing the service on.

For almost any real-world use case, memorizing the password and typing it in on a keyboard is pretty much a necessity. There are many kinds of inputs that don't allow another software like DiceKeys to enter the secret, so I'd still definitely rather stick with the simpler, human-readable option.

You could also just get some random dice, including ones with letters, and generate a password directly from that without going the word-lookup route.

I'm probably an outlier, because I already have random dice, but "right leg squshr december 46 monday" seems secure enough to me: https://i.imgur.com/JGh4fik.jpg

Assuming someone knows you used those dice, but doesn't know the order you arranged them in, and assuming that the d6's with the letters on them are not identical, I think you've got 55 bits of entropy there [1]. That's equivalent to between an 8 character and 9 character password of ASCII non-whitespace printing characters.

Password crack time estimator sites are giving me times ranging from a few months to a few centuries to brute force it. Probably OK but I'd personally want something a bit better for a master password.

[1] 12 x 6^6 x 12 x 6^2 x 7 x 11! / 2. The first 5 factors come from the individual dice, the 11! covers order, and the divide by 2 comes from the two numeric d6 being interchangeable so order does not matter for them.

> Assuming someone knows you used those dice

Under that assumption, yup. But naturally, if you were really going to do this, you wouldn't tell anyone which dice you used, or that you even used dice to generate the password. Ideally you'd probably use even more dice.

Of course, diceware is meant to use something most people can get their hands on easily; a day-of-week die, etc., isn't something you're going to find in a common game set.

The idea is neat, but it seems like a horrendous idea to get your disk encryption keys from something you liked because of the aesthetic!

I guess I see the appeal in that sense. It's a bad practical device, but a good security-themed desk toy. (Though for the latter, I'd prefer a combination Enigma/watch for TOTP?)

BTW, fully agree about that having to trust the website is pretty bad.

Hopefully they'll release open-source scanner ASAP.

Well you don’t have to trust the website, you could fairly easily copy the dice’s contents to your password generating app yourself.

Reply on news.ycombinator.com