bug#37744: Per-user profile directory hijack (CVE-2019-17365 for Nix)

[Tobias Geerinckx-Rice via Bug reports for GNU Guix <bug-guix@gnu.org>]

[-- Attachment #1: Type: text/plain, Size: 1067 bytes --]

Ludo',

Thanks for your answer.

Ludovic Courtès 写道：
>> I need more cluebat please: say I'm an attacker and connect to 
>> your
>> daemon (over TCP, why not), asking it to create an empty
>> ‘per-user/ludo’.
>
> You wouldn’t be able to do that because over TCP because the 
> daemon
> can’t tell what user you are.

No, I ask it nicely: ‘hullo daemon, I'm, er, "ludo"’.

Of course the remote daemon doesn't trust me beyond pre-creating 
an empty per-user directory owned by the local "ludo" user only if 
such a user exists.  It doesn't even report succes or failure to 
avoid leaking valid user names.

You already trust the network not to DoS you with webkitgtks, how 
does this new step decrease security?

Sure, it bumps the protocol version; I'm aware of that.

> It’s meant for cluster setups where you have one
> head node that clients connect to from remote nodes.

And likely some kind of centralised user management so it's not 
unreasonable to handle this differently/manually.

Kind regards,

T G-R

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 832 bytes --]