[numbsafari]

This is a big deal.

If we can find ways to perform secure, multi-party computation, we could develop fully distributed computational, networking, and power delivery systems.

Your solar roof tiles could be, basically, CPUs or GPUs with embedded wireless networking.

[xtreme]

Can you elaborate why solar roof tiles need such intelligence, especially the secure part? I don't see what data they would need to hide.

[IpV8]

Distributed power electronics are inherently dangerous. If you have a surplus of power and your neighbor has a deficit, it is desirable for the two systems to communicate to send power from you to the neighbor. If a bad actor can leverage this system to tell everyone to send all of their power to the grid, they could easily damage the infrastructure in a very costly way.

[infogulch]

Ok so the grid (that knows everything already) commands every house what to do, using a pinned certificate to authenticate. What part of this needs to be zero knowledge?

[lgregg]

It's also important to point out the grid is archaic in many places and the knowledge isn't omnicompetent across the system. In the US, we have three major grids which are also connected to Canada's (which I can't speak too). [0] From my understanding, with an amateur interest, we know a lot about where those major grids connect but very little about the distributed nodes that make up the network; The Northeast Power Blackout of 2003 is a really great example of this. [1] Essentially, there was a grid failure that occurred in Ohio which overlapped into the other sections of the grid. In short, a wire contacted with a tree in Ohio which caused a cascading failure to NYC.

So, let's now bring IoT into the mix. You and I have smart houses, with smart solar tiles. John attacks our tiles plus all our neighbors and directs a major electrical spike towards our local substation. Now it's a physics question, where are all those joules of energy going? It's a heat problem and right now there is no way to dissipate that heat from the system which will melt our substation. Let's say our neighborhood is between several other neighborhoods and the main power station, we just killed a node and guess who else doesn't have access to power because they don't have tiles like we do.

That's the basic concept on why security and authenticity are important in relation to the energy grid. It would be nice if there was an effective way to dissipate an "electrical DDoS"? I'm not sure if it's called something else. If you're interested in the energy dissipation within the energy grid, this is a great question on SE. [2]

That all said, is also a major reason why the government is consistently freaking out about our power grid being hacked.

[0] https://en.wikipedia.org/wiki/Continental_U.S._power_transmi... [1] https://en.wikipedia.org/wiki/Northeast_blackout_of_2003 [2] https://electronics.stackexchange.com/questions/117437/what-...

[davidkuhta]

I'm imagining a world where all the solar roofs have a "Tron Tile" to protect the system.

[numbsafari]

It would create a truly distributed compute utility that was accessible to anyone within range of the network.

Individual tiles could store and compute and cooperate in a mesh network.

To make something like that feasible, it needs to be secure and reliable. In the privately-owned case, I want to ensure that an adversary who gains access to the “grid” can’t exploit it against me. Think of it as having a raspberrypi on the outside of your house where anyone could tamper with it.

In the public scenario, you would want users to be able to offload computational tasks to an ambiantly available computational resource. Some of that computation will need to be local, and for disaster planning reasons, would be helpful if distributed in nature. But, in order for Joe and Jane Schmoe to rely on such a resource, they would want to know that their data isn’t being stolen or shared with others.

[cloud_man]

An innovative project called Golem (golem.network) is working on distributed computational power with data privacy and verification. They seem to be ahead of the pack and think this could be the future of computing.

[petra]

Is secure multi-party computation efficient enough for such use-cases, i.e. shared compute ?