Creating a Debian 13 minimal sys-net with hardening

In the original contents of the topic, I understood “hardening” is toward existing security. The contents might involve some misunderstanding on Qubes OS and bundled templates. But, someone objectively requested evidences. Someone said the security is the same as before. Someone provided check list. So, misunderstanding and errors can be corrected in the thread of objective conversation. (In hard science, objective has a meaning of not referring to people; instead talk like the property of objects are independent of existence of human being.)

What I regret is I missed the timing to propose a procedure like this.

  • Discussion/vote for achievement of alpha-test.
  • Discussion/vote for announcing beta-test. (Beta-test involves users without professional skill of programming. Consequently, further test might be proposed in the process.) I though the topic meant “let’s make the best sys-net” up to here.
  • Beta-test (This might be in the Category “Community Guide” but in a different title chose after discussion.)
  • Evaluation

This procedure has four advantages.

  • Either to confirm security of minimal-template based sys-net or to create sys-net based on minimal-template with some addition.
  • Defend the general user from risk contained in software they install in -xcfe-template. (Users often install software recommended by colleagues. But those software are on vulnerable sites and those software are not signed.)
  • Advertise that the community of Qubes is working for the users.
  • Including more developers in Qubes OS.

As for the fourth advantage: one method of improving documents/products is to rewrite them from scratch time to time and compare with the existing ones. Someone spontaneously doing this method is a good chance. Of course, it might be rare the new one brings an actual improvement. Rather we use this method for finding one direction of improvement out of many trials.

Freshman often have many misunderstanding so that cooperating the volunteer of this method needs patience. Someone people had this patience. I avoided disturbing such efforts.

Another point: some posts connecting the thread might have been lost. I avoid to attack posts in the current state of the topic. Imagine a counter evidence to a post was lost somehow, then, I am afraid my further criticism might be too offensive.

2 Likes

I agree there are many things here that are useful.

Advanced users can here debate. Personally I use many ‘sys-nets’ Open-BSD, this one, a Kali-minimal one, and also a blackarch one… Arguing about which is the best is stupid, everyone has a use-case scenario. I have taken some great tips from here which I extrapolate to others..

I also use many wifi adapters, many usb ones. Also Yagi turbo-tennas etc… I would love to talk about that more in a Qubes Context. What I wanna say is there is no best one, I use everyone differently, look at one of the founding docs from Rootskovia Playing with Qubes Networking for Fun and Profit | The Invisible Things Blog

The beautiful thing about Qubes is we can use many…

Open your minds… Would be great to talk openly about your wireless adventures!

5 Likes

So locking down important security policy, including some default ones, to prevent upstream updates or runtime software from silently resetting those parameters (back to weaker defaults), is a false claim of hardening?

These are already set in the “non-hardened” Debian:

# sysctl -a | grep -f guide.txt 
fs.protected_regular = 2
fs.suid_dumpable = 0
kernel.dmesg_restrict = 1
net.core.somaxconn = 4096
net.ipv4.conf.all.accept_source_route = 0
net.ipv4.conf.default.accept_source_route = 0
net.ipv4.icmp_echo_ignore_all = 0
net.ipv4.icmp_echo_ignore_broadcasts = 1
net.ipv4.icmp_ignore_bogus_error_responses = 1
net.ipv4.tcp_syncookies = 1

If your threat model is silent upstream distro changes through updates, that practically equals distrusting the distro (and Qubes itself). An upstream attacker has practically limitless ways to act through updates.

So, is existing value locking a false claim of hardening in such a threat model? - Yes, and a theatre too.

As for runtime software - well… are you planning to run any malicious user apps in the distrusted sys-net? - If yes, then you don’t need Qubes.

2 Likes

Have you ever heard of Clashing Package Philosophies? Shouldn’t we be able to separate intentional malicious backdoors by Debian developers from predictable day-to-day Linux systems engineering realities?

Furthermore, users frequently follow troubleshooting guides to fix network or driver issues, especially on a minimal template. Is it hard to imagine them accidentally overriding system defaults in the process?

The fact that @tommy mentioned in his thread that some sysctl ‘are optional / already existing’, shows intent of locking down rather than ‘copy/paste’.

Whether you agree or not, hardcoding existing security defaults isn’t a security theater, it is configuration anchoring. A valid hardening point. It may be unnecessary for your use case, but the logic remains sound.

Using labels like ‘false claims’ and ‘security theater’ shuts down productive debate and sends the wrong message.

3 Likes

Shooting yourself in the foot is a different threat model from upstream updates. I answered about the latter.

Using labels like ‘false claims’ and ‘security theater’ shuts down productive debate and sends the wrong message.

I don’t quite understand what you expect a debate to produce, if you keep changing the context of what was actually said.

I don’t know what more to say. A guide like this one is a good idea but needs significantly more work. If you feel you have the network security expertise, improve it. If you like what was written, put it in your sys-net and consume the consequences.

2 Likes

The last time I checked, ‘clashing package philosophies and predictable day-to-day Linux systems engineering realities’ are directly relevant to the threat model of upstream updates. Ask yourself again if i changed the context of what was actually said.

Last time i checked you labeled ‘hardcoding some already existing security defaults’ (sysctl) as a false claim of hardening and theater. Take another look at what was written and ask yourself if I actually changed the context of what was said.

Feel free to stop here unless you spot any actual factual errors in the thread. Let’s avoid labeling differing cybersecurity philosophies as ‘false claims’. Doing so sends the wrong message and suggests that alternative perspectives aren’t welcome on this forum.

5 Likes

We have something in common:

I thought it a good occasion to significantly more work since we found someone with motivation. I also share the feeling of worry about too early transfer of the topic to Community Guide. For the significantly more work, expert’s cooperation is essential.

However, there is another side.

  • Installing a software in a meeting is rather common. Refusing it sometimes means good bye to the community of that user. Some user wants to protect themselves from the software installed in that way. This is one of the motivations for using Qubes OS.
  • Upstream decision of changing some sort of settings occurs often. Some of them might interfere the protection of Qubes. It is a good idea to have some sort of software of Qubes OS community defines necessary parameters. I am using patches in reaction to upstream decisions, though in different branches from security. For security, an interval before the patch is more problematic.

I think it a good thing that user’s side of interest is discussed actively in this community. (Some other things than the contents of the shipped Qubes OS have something to do with security.)

Edit:

I hesitate to call this topic a guide. People were participating in development and verification. Changing the category to Community Guide should be after discussing how it is presented. Something (deleted/moved posts, some posts edited by someone other than the authors of the topic?) makes it hard to what happened. This topic should was moved to Community Guide too early.

2 Likes

if a beginer user followed this guide to every word what wold happen

1 Like

This guide is easy to follow for beginners. It will unlikely cause any issues (when following the steps correctly), as long as you keep a backup sys-net based on your default template so you can switch back if something breaks. The ‘advanced’ warning is just a warning that misconfigurations can happen. So you need to understand what you are doing.

4 Likes

This guide is looking pretty sharp now. Thank you @tommy

6 Likes

Would it be possible to seperate this guide into wifi only steps and ethernet only steps? I have my wifi controller separated into sys-wifi. I could probably figure it out myself but just thought I’d put it out there if anyone was interested in doing so

1 Like

A post was split to a new topic: How to stop network spyware with Qubes OS?

I had a dedicated Fedora minimal template with more than 3 GiB in size for my sys-net. I did these steps in a new Debian minimal and my template is now 1.8 GiB. Cool! I should have had considered Debian earlier.

3 Likes