As part of our quest to decentralize and harden the project, we have switched today to a multi-signature wallet for our Qubes bitcoin fund. This means that no longer can a single person, not even myself, sign an outgoing transaction from our new wallet. For this to happen M out of N signatures is required (we selected N = 13, and M = 6, for the time being). The holders of the keys have been invited from among Qubes developers and supporters from all over the world. Some people might have more than one key, but still fewer than M.
Currently this fund is used exclusively to receive donations, but we’re planning some more usecases for it in the near future.
We have also updated the Qubes Security Pack with information required to verify the authenticity of the address. (We especially encourage anyone making a larger donation to verify the address before sending bitcoin.)
Why is this important? With the Qubes OS userbase growing dynamically, it’s only a matter of time before more powerful actors start to see the project as a target, with the intention to weaken it or shut it down. We want to prepare for that moment as best we can. This involves making the project resistant to wrongdoing by any single person and minimizing the incentive to target core people.
Today’s move to a decentralized multi-signature wallet for the Qubes bitcoin fund is one of the many steps towards achieving this goal. Other solutions we have already implemented include: not-trusting any of the 3rd party infrastructure (e.g. for building packages or storing signing keys), using easy-to-migrate-and-mirror git-based hosting (i.e. not being dependent on, and certainly not trusting, GitHub), Qubes VM-enforced sandboxing for template builds (to protect our developers, and consequently the official builds), using Tor for routing all the traffic to/from the VMs used for building the official packages and releases (to increase the difficulty of targeted injection of backdoored 3rd party packages, e.g. from Fedora or Debian, by an adversary who gets access to 3rd party signing keys), and offering users the ability to easily download all updates over Tor (to increase the difficulty of targeted attacks against our users). And we’re planning more: obligatory multi-signatures on the source code (qubes-builder should require e.g. at least 2 signed tags before building any official package), and, of course, multi-signed binaries and ISOs (for which we need the build process to be reproducible).
In a few years, I hope that Qubes will become fully decentralized, with no single person having the capacity to compromise it.
However, it is important to clarify that by “decentralization” we do not necessarily mean “democratization.” If our goal were to democratize the project, there would be a risk that, without clear leadership, the project could stagnate or wander aimlessly. Even worse, the project could be misdirected into a blind avenue, whether due to lack of skill, naivety, or perhaps the intentional actions of those among whom we decide to distribute power.
Consequently, we’re not interested in decentralization of the current leadership (at least for fundamentally important matters, such as the general roadmap and architecture), but in building in various mechanisms to prevent any potential corruption.