You have to create a super-folder, a Folding@Home Pool miner. If you make a massive simulation and someone is pumping in fake values they will get caught immediately, we could do it in quadrants maybe sections of a massive (n x n x n)n simulation with each node in the chain being a part of the larger simulation, bad values would not match. maybe even partition the jobs into threads so that each node is processing multiple projects at once? This would create multiple faulty computations appear on the larger grid, pinpointing nodes with failing GPUs or CPUs maybe even FPGA chips so that they can be re-assigned at the first sign of trouble.
I have a few ideas on how we could create a massive Hadoop cluster with Machine learning algorithms to properly connect everyone on the grid. Hell with this setup even the proximity of faulty or malicious nodes can be spotted by simply noticing if they are a botnet with multiple failing nodes or a deliberate attack, the machine wouldn't have to figure it out, it would just disconnect them from the grid.
The more Vectors we have to spot a faulty computation the better.
(n x n x n)n Grid to overlap computations.
Partitioned computations so no single node gets a single job, further dividing the possibility of damaging a major part of a computation.
Further partitioning through time as well can make the system more robust, you get to only complete a set size of a computation before you are handed a new one... further complexifying this Rubick's Cube. The system optimizations have to be automated.
Adding Machine Learning Algorithms to Data Mine all the connections for Performance data; This data will be important, it tells the system what a normal working system looks like, deviations are suspect and will be investigated with additional MLAs. This will be useful to analyze Folding Algorithms, Network Connections and how to route nodes for fastest performance, even further analysis metrics that can be open sourced so that anyone can view the metrics of the Grid to add that Human Intelligence component, we are masters at pattern recognition why not integrate it into it's working; The Machine will learn from us on how to best operate the network, by simply adding suggestions and giving the Network time to analyze controlled changes in it's parameters... a massive crowd sourced control room, maybe even a way for people to learn how the science of folding works so they can spot bad computations? The two systems would be separate, but the system could also be helped to compute the results to see if crowd sourced suggestions would have helped it deal with it's network problems for future use.
I believe in using every ounce of intelligence we can, and that requires a little more than computers, Each processor is suited to different computations just have to build it in a way that is accessible to all the architectures, even our own.
The back end for the points can be tied in directly with the Open Transactions Library for security of the point system, it would be tamper proof, all nodes would be servers for the points network, making any change would still require the concensus of everyone but without a massive blockchain. It could be implemented by writing a proper OT contract on how points are to be handled per computation and simply automate the dissemination directly from the OT server federation... the contract would be impossible to tamper with.
The contract made would give authority of the points calculated for a set time for the network to do the Data Mining, once done making sure everything is normal, it releases the points automatically.
Any further action if the calculation is proven to be faulty at a later time, we can give them a demerit point and work under the philosophy of, "Fool me once, shame on you! Fool me twice, shame on me."
A lot of the system could be directly accessible from the web if we convert it to JavaScript, a special subset asm.js, to deploy it across the web for all processing devices, so that even a cell phone can donate it's metric data for science.
here is an example so you can be amazed... all this is JavaScript. You will need Firefox
http://www.mozilla.org/en-US/firefox/new/Epic Citadel: Unreal Engine 3 Port to asm.js
http://www.unrealengine.com/html5/BananaBread playable demo.
https://developer.mozilla.org/en/demos/detail/bananabreadMonster Madness game: Zombie asm.js port, test servers won't be back till january , so bookmark it for later.
http://www.monstermadness.com/OpenRisc emulator with Internet access... it's old so use the console.
http://s-macke.github.io/jor1k/and more demos.
https://github.com/kripken/emscripten/wikiHow hard could it be to implement all the security and folding at home software into a browser for instant access? We can even make it fully opensource so that it gets the input of thousands of dedicated volunteers, they will make the code unbreakable, if they can't break it by knowing how it's built, then no one can and when it's ready it will released for everyone to try to break, before integrating it permanently.
