top | item 43938835

(no title)

thecrash | 9 months ago

Your problem statement is effectively "I want to share access to my documents very informally with people who don't care to have any security practices, but still keep them secure"

There's another way of sharing in cryptpad though, which is for each user to create an identity/account. Once those you're collaborating with have accounts, documents and folders can be shared by granting access within cryptpad's UI. No secrets have to be circulated.

discuss

order

jraph|9 months ago

> Your problem statement is effectively "I want to share access to my documents very informally with people who don't care to have any security practices, but still keep them secure"

Well, that's certainly what tools like CryptPad and Signal target: privacy for the non expert.

OP' points are right, and I hope they get addressed at some point.

hmsimha|9 months ago

I've worked with a few orgs which have used cryptpad, and I'm sorry, but Cryptpad doesn't make it possible to share documents securely unless again, everyone in the org is able to follow security protocol to an exceptionally rare degree.

Even you seem to think sharing via identity somehow bypasses the problem, when in fact this just sends them a "notification" with a share link containing the same secret URL (not to mention, as far as I can tell, there's no way for them to add the document to their own drive, so if they want to access it later they either need to save the share link or find it in their notification panel under "notification history" which is super unintuitive).

And again, those secrets are stored in your browser history. In a group I was involved with, the workflow was to create documents and share them with others, or put the share link in a Signal group. Even if one were to try to tell everyone in the group that the link should only be opened in a browser that doesn't share its history with its vendor, clicking the link in Signal will happily just open it in which ever browser is configured as your system default anyway.

Cryptpad effectively gives you the rope and then ties it into the noose around your neck for you, while you're not looking.

Security theater is at best mildly dangerous in a more typical scenario where it's constructed around an application that isn't billed as a secure communication platform. When a tool advertised as user-friendly and privacy-enhancing is the subject of such theatrics, it's even more actively harmful because it instills a false sense of confidence. It would be like a safety helmet that explodes when the user grazes their head.

So to recap, if you care about big tech companies gaining access to your secure documents, the only way to use cryptpad in a remotely secure manner, in a group, is either by password protecting all documents with a strong password, or ensuring no one in your org ever opens a document in a browser with history syncing. And honestly, expecting the latter from 99% of groups that might use cryptpad is unreasonable, which is why I'm saying it's irresponsible of Cryptpad to even allow password-less document creation (without so much as showing users a glaring red danger notice beforehand).

The users are not primarily to blame for incorrect use of a software that's billed as privacy-preserving, when that software drops them off at the happy path and neglects to tell them, "by the way, we've booby-trapped the door to fire a footgun when opened unless you also turn the smaller knob on the far side with your other hand as you open it."

I realize the data exfiltration issues I mentioned are non-trivial to address (though by no means an immense project either), but I can't interpret the link situation as anything other than willful negligence, or worse, a honeypot; consider that users whose adversaries might include nation-state actors (for example, undocumented immigrants sharing resources with one another on how to access services while staying under the radar) are perhaps more exposed, because data brokers are more likely to deny state requests for data that can be seen as overly broad, whereas one specific type of data (browser history) on one domain becomes a pretty tightly scoped request.

ldubost|9 months ago

Your last paragraph is quite insulting to the work we do, suggesting intention to trap people ? Did I read this right ?

I'm not really sure i want to continue the conversation unless you retract this. Our team is working hard on many fronts and does not deserve to be treated like that.

If you believe it's critical that the "link situation" be resolved, where is the pull request, or even the specification of the necessary change ?

Ludovic