Posting that I was wondering that people like structured data more than random… x4 postal addresses are easier than x13 words… but it’s still not ideal and lazy for everyone’s interest.
At some point if that old idea of walking up to a Safe Network terminal anywhere, took off, perhaps that would see a hardware token take out the hard work… but trusting hardware is a problem unless there are already usb’s that control what can occur.
I don’t know if it’s possible to turn on its head and allow uses to choose ~13 words, from that book on the shelf that is not accessible to others… but I wonder that’s known to be a bad idea for allowing guessing. Mixed perhaps with switching out a couple of words??..
Therein lies the issue. We need random, and humans are bad at random, try as they might.
If you are aiming for memorable, then it’s likely going to be a sentence (a non random phrase) and if it’s a human choosing it it’s likely to be not so random. It’d probably end up being a well known bible verse or Shakespeare quote, and therefor problematic.
But if memorable isn’t a requirement, then why not just print the random Passphrase and hide it in the book
On the subject of making wordlists that produce phrases that are more memorable, check out this blog post, particularly the bit at the end about using 2-grams and 3-grams from very large word lists, which might be closer to what you are thinking.
Four 2-word phrases would produce 120-bitsw of entropy, which is close to what we need. However, it would be slower to enter, and off the bat would have no checksum and only in a single language etc. So trade-offs!
You only said that cos you dont like being called turntable
These were far too obvious for anyone closely connected with the project of course but if 4x2 word phrases are all we need or 4 of say 7 two word phrases then I think we ( or rather you) have a winner.
Well done Jim
… and we’re using the Safe Network
which is pure random.
If only you knew where to look…
If user can recall a location they know that is a place that the network holds a longer string?.. but not just one a few perhaps is enough to stop someone else… coupled with xyz??
Yeah, they need to be randomly generated by the client to be in any way secure (see the brain wallet attack link to see why it’s a bad idea to have a go yourself).
I’m not advocating we go for that scheme BTW, as I think it has numerous drawbacks (and a ton of work) as opposed to BIP39 wordlists, which will be usable off the shelf in short order. Pragmatic approach I think.
Ah! I’m talking specifically about the Passphrase part here, which at a minimum we’d be wanting 128-bits.
Then in combination with the Password we can easily get over 160-bits (even with very lenient password requirements: 8 characters required) which we’re considering the bare minimum. With device keys and other combination of keys, the sky’s the limit.
But, if we did want to get to 256-bits with just the Password + Passphrase combo from the off (which may be overkill) it’s not difficult to do:
10 characters including uppercase and numbers, in the Password
Extend the Passphrase from 12 to 18 words
And boom! 256-bits.
I did have it in mind that during on-boarding we could do a strength calculation on the users chosen Password, and then tune the Passphrase length to meet a minimum entropy. Or perhaps give the user some feedback and options to increase strength, but for the purposes of demonstrating the concept, and not getting too far into the weeds on these wireframes, I’ve not included that.
I would like to bump this question. As I think it has not been answered yet.
There is a chicken/egg problem right here. Presumably you would need safe tokens to store your created/registered account to the network. As otherwise it would be trivially easy to take the network down by spamming free accounts.
So we need to transfer x amount of safe tokens to an account that doesnt exist yet, so it can be created.
As far as I understand it, this is fixed by Passphrase&||Passwords being only stored locally. (more like they are irrelevant as long as you don’t put anything into the network, as only way they arrive at the network is encrypting something)
Still settings like 2-of-3 makes me question my assumption. So I would like some more clarity here. What parts of these process will require safetokens & what parts are stored/handled only locally.
Because I hope it will be impossible to put ANY data to network for free. Because it would definitely be abused.
Sorry, I missed this Q at the time. You’ll be pleased to know this isn’t an issue.
The Safe ‘creation’ process happens offline. In fact, you don’t need to put your Safe on the Network at all if you don’t want too. You can load it up with data, and it’ll be locally encrypted on your device (so you get some security benefits right off the bat).
But when you are ready to add this data to the Network—say for example you wanted to sync across devices, or publish some data—then you’ll get a quote for some or all of the data in your Safe, and can earn/purchase/access the required tokens for that, drop them in your Safe, and off you go.
You’ll get feedback when the data is safely stored on the Network, and can be locally deleted etc.
Just to clarify, there are no “accounts” on the network in the traditional sense, it’s just keys. A Safe is a way of you controlling a set of keys, and generating them, and accessing data via them, but in a way that is more straightforward to spatially conceptualise, and therefore be in better command of.
You are correct. It won’t be possible to get the Network to store any of your data without paying for it.
I hadn’t picked up on this Jim, and maybe I’m not the only one! Very cool.
Do you have the CLI or any app UI sketched out for how this would look to the user? And how the local and network locations for data will be apparent and managed?
Also, does this mean that your devices each need a local (encrypted) copy of your Safe or will you be able to see everything that’s in your Safe without necessarily having it all stored locally?
Imo, it is pretty cool that you can have a local copy. It would be quite Dropbox like in that respect and there are obvious performance advantages from have a local copy too.
Obviously, for sensitive data which needs to leave no trace on a local device, there is a different perspective. The average user may prefer performance over absolute privacy though, imo.
Well if its the account to view public files then it prob doesn’t matter too much since no private data other than playlists and small amount of token held
No not yet! I have some ideas, but taking it one step at a time, and that started with the Safe creation process.
Yes, absolutely. You’ll be able to see what’s local, what is safely stored on the Network, and what is pending local changes etc.
Yeah, that’s my feeling. CRDT should allow us to do some pretty useful things here.
There will be a variety of options for people depending on their requirements. But yeah, having performance for editing with apps locally, and then incremental changes syncing to the Network in the background as you go, will be pretty useful.
Some people may want zero trace of data on their local device in some circumstances, and that’s cool but also remember the idea would be that anything that is dropped in the Safe gets chucked and encrypted locally before is goes unto the Network, so you’d be benefitting from that even with an offline/air-gapped device.
(Worth remembering that in theory you could have a useful Safe and use Safe Apps without ever writing to the Network… or perhaps never even have your device on the internet at all!)
The other benefit to having things deleted locally when they are safely synced to the Network though is piece of mind that a lost device couldn’t be brute forced (even though that’d extraordinarily tough!) because I can nuke it’s keys remotely once I realise it’s missing.
