!ZRgXNaHrdpGqwUnGnj:nixos.org

NixOS Security Triage

762 Members
Coordination and triage of security issues in nixpkgs234 Servers

Load older messages


SenderMessageTime
30 Jul 2021
@linus.heckemann:matrix.mayflower.deLinux HackermanThe hardened profile breaks things. Don't use it if it break things you need.14:39:22
@linus.heckemann:matrix.mayflower.deLinux HackermanSorry, I'm in a bit of a foul mood for unrelated reasons and this probably isn't the best disposition to be helping people in. I'll be off.14:40:27
@tnias:stratum0.orgtniasEspecially on a desktop/workstation it is not usable. Would not recommend.14:41:14
@philipp:xndr.dephilipp changed their profile picture.20:21:15
@mlieberman85:matrix.orgMichael LiebermanHas anyone taken a look at OpenSSF's new SLSA standard for supply chain security? https://slsa.dev/ Seems like just by its nature Nix hits the highest level for most things. https://slsa.dev/21:00:55
@sandro:supersandro.deSandroDefinitely not A22:45:29
@sandro:supersandro.deSandroAnd most of the other could be circumvented but not in the default configuration. Also we could have bad npm packages, too.22:46:19
@sandro:supersandro.deSandroBut we avoid some attack vectors due to the build sandbox22:46:32
@mlieberman85:matrix.orgMichael LiebermanInteresting. I'm not super deep yet on Nix internals. Does SLSA seem like a reasonable standard? It's pretty new and based on Google's internal Binary Authorization standards. I'm a bit cautious on some of the claims the SLSA standard makes because it relies on "trusted control plane" and similar. If so is there any doc or anything I could read up on regarding current Nix security concerns (that aren't confidential)22:53:46
@sandro:supersandro.deSandro I don't know if it is reasonable but I didn't read to much weird stuff yet. 23:09:03
@sandro:supersandro.deSandroNix has probably all of the security problems that come with a big open source project where not everyone knows everyone and every part of the code23:09:56
31 Jul 2021
@roosemberth:orbstheorem.chRoos

Dependencies have their own SLSA ratings, and it is possible for a SLSA 4 artifact to be built from SLSA 0 dependencies

Specially this.

08:30:08
@roosemberth:orbstheorem.chRoos We may have provenance, build signature and somewhat reproducible builds (arguable), but we're still pulling stuff from unknown sources. 08:31:07
@sandro:supersandro.deSandroYeah well, we need to get the source from somewhere08:33:14
@roosemberth:orbstheorem.chRoosIMHO, SLSA 2 is missing non-repudiability.08:33:36
@roosemberth:orbstheorem.chRoos
In reply to @sandro:supersandro.de
Yeah well, we need to get the source from somewhere
Yes. Security-sensitive processes do review source changes before using them, we don't.
08:34:36
@disrupt_the_flow:matrix.orgdisrupt_the_flow
In reply to @linus.heckemann:matrix.mayflower.de
The hardened profile breaks things. Don't use it if it break things you need.
Yeah I know and I fixed some but this specific one is weird.
08:35:03
@roosemberth:orbstheorem.chRoosInteresting read, thanks ^^08:36:49
@sandro:supersandro.deSandro
In reply to @roosemberth:orbstheorem.ch
Yes. Security-sensitive processes do review source changes before using them, we don't.
I am pretty sure security sensitive processes also try to use as little packages as possible and not literally anything. I think we do it sometimes for core packages but not for every package
08:37:34
@roosemberth:orbstheorem.chRoosOh, I didn't know we did source-review!08:38:31
@r_i_s:matrix.orgris_.... depends what you mean by source review ....18:51:54
@r_i_s:matrix.orgris_and what sort of attack scenario we'd be trying to catch by such a review18:52:31
@r_i_s:matrix.orgris_there are few if any packages where we review the (source) diff of every bump18:54:20
@r_i_s:matrix.orgris_and i'm not sure there are any distros that do18:54:38
@r_i_s:matrix.orgris_anything other than an extremely minimal distro trying to do that would get so bound down in molasses that i would imagine any security benefits from "supply chain security" would be outweighed by the slowness of it all.18:55:59
@r_i_s:matrix.orgris_anyway...18:56:04
@r_i_s:matrix.orgris_(was going to go on for some rabbitmq/elixir assistance but i think i've figured it out19:15:24
@r_i_s:matrix.orgris_ * (was going to go on for some rabbitmq/elixir assistance but i think i've figured it out)19:15:29
@r_i_s:matrix.orgris_actually it does look like i'll need to call in some rabbitmq help on https://github.com/NixOS/nixpkgs/pull/13224221:46:24
@mlieberman85:matrix.orgMichael Lieberman
In reply to @r_i_s:matrix.org
anything other than an extremely minimal distro trying to do that would get so bound down in molasses that i would imagine any security benefits from "supply chain security" would be outweighed by the slowness of it all.
I think it’s a balance. Not everything needs to be slsa4. And you can be slsa4 for your source and build but include slsa0 dependencies
21:54:15

Show newer messages


Back to Room ListRoom Version: 6