Bitcoin Forum
November 12, 2024, 02:30:27 PM *
News: Check out the artwork 1Dq created to commemorate this forum's 15th anniversary
 
   Home   Help Search Login Register More  
Pages: « 1 2 [3]  All
  Print  
Author Topic: MerkleWeb - statistical Godel-like secure-but-not-perfect global Turing Machine  (Read 12229 times)
phillipsjk
Legendary
*
Offline Offline

Activity: 1008
Merit: 1001

Let the chips fall where they may.


View Profile WWW
December 27, 2011, 05:00:48 AM
Last edit: December 27, 2011, 05:35:08 AM by phillipsjk
 #41

...the faster algorithm only has to be written once (unless encrypted).

Wait a minute...You mean the network can run programs without knowing what they do?
I don't get it. Can you explain it?


It is possible that I made a slight cognitive leap based on my limited understanding of Homomorphic Encryption.

In practice, (depressingly) few file formats clearly distinguish between executable code and data. Often due to programmer laziness or feature creep from management[Citation Needed].

I was making the apparently false assumption that the key-renewal algortihm described in the Damien Stehle and Ron Steinfeld paper was Turing-complete and encrypted. However, skimming the paper without understanding most of the math, it looks like the renewal algorithm is known, but additional "hints" are stored with the public key. This extra information is used to reencrypt/error-correct the private key stored with the encrypted data.

Edit: the one time I don't check links... or were you concerned by the lack of peer review?

James' OpenPGP public key fingerprint: EB14 9E5B F80C 1F2D 3EBE  0A2F B3DE 81FF 7B9D 5160
netrin
Sr. Member
****
Offline Offline

Activity: 322
Merit: 251


FirstBits: 168Bc


View Profile
December 27, 2011, 05:30:16 AM
 #42

Your link is fishy, try this: http://eprint.iacr.org/

Also: http://css.csail.mit.edu/cryptdb/

Quote
CryptDB ... manages to emulate fully homomorphic encryption for most of the functions of the SQL databases used in many applications, computing only with encrypted data and adding just 15% to 26% to those applications’ computing time. ... RSA scheme, for instance, lets you multiply encrypted numbers, and another called the Paillier scheme lets you add them. ... plenty of schemes allow you to compare encrypted information and see whether the scrambled codes are equal ... SQL queries in a database are composed of relatively few types of operations: equal to, less than, summing up, sorting ... we were able to find an encryption scheme that is quite efficient at computing on encrypted data. ... switch between crypto systems on the fly depending on the operation. The data in the database is encrypted in multiple layers of different encryption, what the researchers call an “onion” of encryption ... it does “leak” information about the underlying data when enough outer layers of encryption are removed

--- Andy Greenberg, Forbes

Greenlandic tupilak. Hand carved, traditional cursed bone figures. Sorry, polar bear, walrus and human remains not available for export.
phillipsjk
Legendary
*
Offline Offline

Activity: 1008
Merit: 1001

Let the chips fall where they may.


View Profile WWW
December 27, 2011, 05:43:20 AM
 #43

What I think what jtimon getting at using the Socratic method is that the data is secret, not the algorithm. That probably has all kinds of interesting implications for data leakage or separating data and executable code.

James' OpenPGP public key fingerprint: EB14 9E5B F80C 1F2D 3EBE  0A2F B3DE 81FF 7B9D 5160
jtimon
Legendary
*
Offline Offline

Activity: 1372
Merit: 1002


View Profile WWW
December 28, 2011, 08:09:57 PM
 #44

I think I now get what you mean (although I'm an homomorphic encryption newbie).

And, yes, I guess you could transfer part of the algorithm to data. For example, a trick that comes to mind...
You could train a neural network to be equivalent to a given function or part of an algorithm, then use the weighs of the network as part of the input (although constant) and be able to also encrypt them, obscuring the algorithm.
Probably not very attractive for merkleweb if the number of inputs are a limitation.

By the way, can someone explain me how homomorphic encryption can be used to "create secure voting systems"?

2 different forms of free-money: Freicoin (free of basic interest because it's perishable), Mutual credit (no interest because it's abundant)
phillipsjk
Legendary
*
Offline Offline

Activity: 1008
Merit: 1001

Let the chips fall where they may.


View Profile WWW
December 29, 2011, 06:17:03 AM
 #45

I don't know much about homomorphic encryption either.


By the way, can someone explain me how homomorphic encryption can be used to "create secure voting systems"?


That claim in the Wikipedia article was cited:

Quote from: Wikipedia

Even if it does work as described, I would be concerned by the lack of paper-trail. It is very hard to trust that a computer is doing what it says it is doing. Even if the vote tabulation is beyond reproach (thanks to the proposed distributed turing machine), the computer being used for voting may be able to do statistical vote substitution (even in the face of vote reciepts).

James' OpenPGP public key fingerprint: EB14 9E5B F80C 1F2D 3EBE  0A2F B3DE 81FF 7B9D 5160
jtimon
Legendary
*
Offline Offline

Activity: 1372
Merit: 1002


View Profile WWW
December 29, 2011, 11:47:46 AM
Last edit: December 29, 2011, 12:38:54 PM by jtimon
 #46

That claim in the Wikipedia article was cited:
Quote from: Wikipedia
Thank you.

2 different forms of free-money: Freicoin (free of basic interest because it's perishable), Mutual credit (no interest because it's abundant)
BenRayfield (OP)
Sr. Member
****
Offline Offline

Activity: 316
Merit: 250



View Profile
January 11, 2012, 06:30:37 AM
Last edit: January 11, 2012, 07:06:13 AM by BenRayfield
 #47

(I'll respond to the last comments later)

Me and James Jaeger (director of Fiat Empire - The Federal Reserve Is Unconstitutional and other documentary films, which you can watch on Youtube) are talking about forming a business together to create "IP Walkie Talkies", a wireless mesh network of cell-phone like devices that form a free peer to peer Internet thats off the grid, no monthly fees, no central control, freedom of speech, and it may have some similarities to MerkleWeb. We're still talking about how to build it. Soon we may be looking for employees and business partners. Anyone interested? Join the talk here:

http://www.kurzweilai.net/forums/topic/a-technological-renaissance-peer-to-peer-non-internet-ip-walkie-talkie-nets

We have the right to freedom of speech therefore also have the right to create and use communication systems which are free of anyone controlling that communication.

We could put every Internet Service Provider and Cell Phone business, out of business, since we would undercut their prices at ZERO per month and no censoring or central control. Don't worry about us becoming a monopoly, since my condition of helping with the project is it is toward an open source hardware and software, but not necessarily starting that way. Also it will prove to everyone that peer to peer Internet can work on a large scale, so others will start work on other forms of peer to peer large scale communication even if something goes wrong with that plan.

Its not hard to obsolete an infrastructure that's designed to crawl slowly as it waits for permission from a hierarchy organization. Peer to peer is naturally faster. For example, Apache Cassandra has a mostly peer to peer core and is used by Netflix to stream their videos, but peer to peer is only possible in server connections, while most Internet connections are client and can't connect directly to eachother (a form of censoring). Peer to peer is the future. Want to free the Internet and get rich at the same time?

Technical estimation, quoted from that thread:
Quote
The technical parts can be done, period. Here's why:

The following is true of IP Walkie Talkies and the existing Internet. Its a property of communication through space, not any specific communication system unless it can communicate faster than light. The existing Internet can be modeled using similar equations by counting the Internet Backbone as a large number of connected devices. These calculations are an estimate of what we need, to be calculated more exactly later. Not all of the variables below are used. Some of them are there for us to talk about later in this thread, to have standard words for these things.

networkArea = flat area the devices cover.

widthOfNetwork = squareRoot(networkArea). We're not going to consider the corners yet. Its approximate.

deviceQuantity = how many devices spread evenly in networkArea.

deviceWidthOfNetwork = squareRoot(deviceQuantity)

distanceBetweenDevices = widthOfNetwork / deviceWidthOfNetwork

radioDistance = how far the devices can reliably transmit a signal to other devices.

radioHopsWidthOfNetwork = widthOfNetwork / radioDistance

inefficiencyMultiplier = how many times more bandwidth than the perfect straight line path is actually used.

connectionsPerDevice = how many near devices each device communicates with.

bandwidthMultiplierForLongestPath = radioHopsWidthOfNetwork*inefficiencyMultiplier. This means how many times more bandwidth is used to send some bytes from one side of the network to the other, relative to how much bandwidth each device uses to do that together.

averagePathLength = average distance between endpoints where 2 devices communicate to eachother, as people command the devices to do.

bandwidthMultiplierForAveragePath = bandwidthMultiplierForLongestPath * averagePathLength / widthOfNetwork

maxBitsPerSecondPerDevice = bandwidth of each device. This is not about how much bandwidth it can actually use in the network.

fractionOfTimePeopleActivelyUseIt = People don't use their Internet connections or phones all the time. Even when reading a webpage, the computer is mostly idle. It uses the connection when you click or stream. This is about the times people normally use it, not in the middle of the night or early morning. Sitting at the computer isn't using it. Actively downloading or talking is using it. People will be motivated not to download constantly because the algorithms will give you only as much bandwidth as your hardware puts in.

averageInternetSpeedBitsPerSecondPerDevice = maxBitsPerSecondPerDevice / bandwidthMultiplierForAveragePath

activeInternetSpeedBitsPerSecondPerDevice = averageInternetSpeedBitsPerSecondPerDevice / fractionOfTimePeopleActivelyUseIt. This is how much speed people will actually see in the IP Walkie Talkie Internet and voice communication system, where voice is transmitted digitally as bits so its simply a realtime Internet at the core.

bitsPerSecondForAudio = how many bits per second used for people talking to eachother through the IP Walkie Talkies.

Example:

networkArea = USA is 9,826,675 square kilometers. http://en.wikipedia.org/wiki/USA

widthOfNetwork = 3135 kilometers. Remember these calculations are approximate.

radioDistance = 5 kilometers.

radioHopsWidthOfNetwork = 627 hops.

inefficiencyMultiplier = 4, since its not going to find the exact shortest path, and we need some redundancy so theres no single point of failure. A single point of failure is often ok in a hierarchy based Internet since theres much fewer and each longer hops, like 20 hops, but not in a peer to peer system which must have many more and shorter faster hops.

bandwidthMultiplierForLongestPath = 2508

averagePathLength = 1000 kilometers.

bandwidthMultiplierForAveragePath = 800.

maxBitsPerSecondPerDevice = 8,000,000. This is 1 megabyte per second. Most network cards advertise at least 100 megabits (12.5 megabytes) per second. We're estimating lower because we have to do it on rechargable batteries and longer distance to the next device.

averageInternetSpeedBitsPerSecondPerDevice = 10,000

fractionOfTimePeopleActivelyUseIt = 1/20. Remember this is only when you click or stream, not when you're just reading a webpage.

activeInternetSpeedBitsPerSecondPerDevice = 200,000 bits per second.

That is 25 kilobytes per second. CD quality audio is 88 kilobytes/second. 25 is more than enough for phone audio (very long range Walkie Talkies, not necessarily connecting into the phone network, and not providing 911 service) and is also enough for very low quality streaming video and to load most webpages in a few seconds. Don't forget there is no monthly fee. Just buy the IP Walkie Talkie and plug its USB wire into your computer for free Internet service for life, and talk to people across the country with no monthly fee. The 25 kilobytes per second (if the research finds the radioDistance and other numbers match this example) is paid for by your device contributing to the wireless network while you're not using it. Its slower than "high speed Internet" but as technology advances this freedom Internet and voice communication will be faster.

...

Constant data, like Bittorrent files or data on the Bitcoin network, could be done a different and much more efficient way, especially if we included hardware to quickly calculate a few common secure-hash algorithms, but those improvements for specific types of content can be done later.

jtimon
Legendary
*
Offline Offline

Activity: 1372
Merit: 1002


View Profile WWW
January 11, 2012, 07:30:21 AM
 #48

Check this out, maybe interests you:

http://www.servalproject.org/

2 different forms of free-money: Freicoin (free of basic interest because it's perishable), Mutual credit (no interest because it's abundant)
phillipsjk
Legendary
*
Offline Offline

Activity: 1008
Merit: 1001

Let the chips fall where they may.


View Profile WWW
January 12, 2012, 10:36:50 AM
 #49


I will respond here for the moment because I don't feel like registering yet another password to keep track of.

My most imortant question is: how essential is the walkie-talkie format?

As Asimov1 pointed out, you are going to find yourself duplicating the existing cell-phone network in many ways.

I can confirm that it will be difficult to broadcast 10 miles. CB radios can do it, but they use about 4 watts of power. The USB interface supplies only 2.5 Watts.

Highly directional antennas will only work on a base station (read: not constantly moving around); also known as a "radio tower" if mounted above the ground for good coverage.

Patents came up. The 802.11n standard uses spatial-division multiplexing. Multiple antennas are driven slightly out-of-phase (glossing over details I don't know alert) allowing the beam pattern to be "steered" somewhat (all without moving parts). Any radio technology from the last 20 years is likely dripping with patents.

Finding spectrum will be a problem. During the last sectrum auction in Canada, the incumbents doubled their ammortized infrastructure costs by bidding something stupid like $4 Billion. They expect to make that back by charging the average cell-phone user $40/month. You will have trouble getting spectrum cheaper because the incumbents won't sell spectrum for less than they paid at auction. So, you are limited to the very limited "free for all" spectrum slices. These typically restrict transmit power, such that your range is likely to be closer to 1 mile than 10. Using the visible spectrum is a possibility for line-of-sight hops.

Finally, I would advise you to not mix the merkleweb proposal with the IP-Walkietalkie proposal. Both are complicated enough on their own that the chance of failure is high. If both projects are successful, you can combine them later.

James' OpenPGP public key fingerprint: EB14 9E5B F80C 1F2D 3EBE  0A2F B3DE 81FF 7B9D 5160
Pages: « 1 2 [3]  All
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!