I’m currently working on my bachelor’s thesis, which involves creating a Qubes OS-based workstation for system administrators. The goal is to leverage Qubes OS to build a secure yet practical environment for daily admin tasks. I’d love to gather your thoughts and brainstorm together. Here are some initial ideas and questions:
Domain / VM Structure in Qubes OS
How would you organize and split up the domains or VMs?
Would you create a separate VM with internal access to the company network?
Would you keep management consoles in their own dedicated VM to isolate critical admin tasks?
Or would you prefer multiple smaller VMs, each for a specific task (like monitoring, configuration, user management, etc.)?
Daily Workflow & Tooling
Which tools and software would you run in which domain/VM?
Do you see any value in having dedicated Disposable VMs for riskier tasks (like testing unfamiliar software, fetching updates, etc.)?
How would you transfer files or logs between domains? Any best practices to avoid cross-contamination?
What are the must-have admin tools in your daily routine?
Security vs. Usability
Which security measures are absolutely essential?
Where might you make trade-offs for a more user-friendly experience, even if it slightly increases risk?
How do you ensure that daily tasks don’t become too cumbersome (e.g., constantly switching domains or copying files)?
Additional Considerations
How do you feel about network segmentation within Qubes (e.g., a dedicated firewall VM)?
Any experience with automated patch management in a Qubes environment?
What should we keep in mind for scalability (e.g., when there’s a larger admin team or broader responsibilities)?
Any best practices you’d recommend when working with Qubes OS?
I really appreciate any insights, tips, or experiences you might share—everything helps me refine my project. Thank you in advance for your feedback!
I do sysadmin for clients and myself, here are how I proceed:
For each client, I have
a sys-vpn qube as they all require a VPN
a sys-vpn-firewall because you need a safe netvm between qubes and the sys-vpn for holding the firewall rules
a ssh qube
a vault qube
desktop stuff qube (conferencing, emails)
I only allow clipboard to work between the qube of a given clients, so I can’t inject crap or exfiltrate data accidently.
This highly depend the trust I have in source material, I rarely use disposable except if I need to try something weird that could damage the user profile in a qube.
Tools I use are usually those:
firefox or chromium depending what I need
keepassxc for the vault and ssh agent
nix, terraform, git, toolschains (nodejs, rust, gcc, go), pip, npm, whatever I need for a task
gimp to handle pictures
I always ensure the vpn firewall does not allow to bypass the VPN, so I’m quite sure I won’t leak client data outside the VPN.
I make sure the clipboard is limited, and that the backup policy matches the one at my client, most of the time I do not backup my clients qubes by default, I just do a backup into the qube itself to send it to my client’s storage, otherwise it could be considered as data exfiltration.
I don’t split the tasks into more qubes, I need to be productive and the security level reached is already far superior to what most of my clients would hope for themselves.
As usability tradeoff, I never reboot qubes after a template update as I shutdown my computer every day, this is way less disruptive.
it is simply a dedicated qube to connect to remote hosts using SSH, and it’s using split ssh for the key as I store it in the Vault of that client.
There is a dedicated tab in “Qubes global manager” GUI program. I need to check how to do the same using tags and write the policy myself, but the GUI works fine although it takes 5 minutes to configure a new client.
I use qvm-firewall to add rules for the client VPN, rules are applied on that qube’s netvm which is sys-firewall in this case. The rule is just deny everything except the host and port required to establish the VPN.
yes when I finish my day, I just shutdown the computer.
Hi ! I’m interested by the discussion and I would like to inquire some more information about your network setup.
Why the client-vpn-firewall ? Wouldn’t setting a firewall rule for the client-vpn be enough ? As rules are enforced in the sys-firewall Qube (or am I remembering things wrong?).
ex:
qube-client → sys-client-vpn → sys-firewall (drop any packet from sys-client-vpn that try to connect elsewhere) → sys-net
Firewall rules of a qube are done on that qube’s netvm. As sys-vpn is on the network border (like sys-net) due to the VPN, you should not really trust it to hold firewall rules for the qubes below. That’s why you need a sane firewall qube between a vpn qube and AppVMs below.
