Artificial neural network VS Bitcoin Address.

mikeywith

legendary

Activity: 2436

Merit: 6643

be constructive or S.T.F.U

Quote from: seoincorporation on December 30, 2018, 12:06:33 AM

Hey mikeywith great answer!, now i have clearly why this is impossible. The way you explain it (with apples and sticks) is clear for any one. So, we can say the score is Bitcoin 1 - ANN 0.

Actually it's more like SHA256 1 - ANN 0 . we need to give credit to Merkle whom i think was the first person to introduce collision-resistant hash function.

seoincorporation

legendary

Activity: 3346

Merit: 3130

Quote from: mikeywith on December 29, 2018, 07:22:12 PM

Quote from: seoincorporation on December 18, 2018, 02:55:54 PM

So, the rules were: give a table of data with 3 colums. first the temperature, then the pressure and at last the property we want to get, it could be entropy, density, viscosity and other stuff... Run the ANN ad if we get a R close to 1 (linear regression coefficient), then our equation was right.

in theory you can, but practically you just can't. the difference between the example you gave is that that first both input and output are small, and the result does not have to be 100% accurate, also the output can be reversed with a few possibilities.

for example if you are only allowed to multiply 2 numbers as an input to generate a single output

if you have

3 * 4 = 12

if you have

3 * ? = 12 > easy to solve the second input

if you have

? * ? = 12

you have a few options to try out

1* 12 = 12

2 * 6 = 12

3 * 4 =12

4* 3 = 12
----------------------------------

but doing this with a 256 bits you always have 2^256 combinations !

the strength of the hashing function is that you can't know where it came from.

Hey mikeywith great answer!, now i have clearly why this is impossible. The way you explain it (with apples and sticks) is clear for any one. So, we can say the score is Bitcoin 1 - ANN 0.

mikeywith

legendary

Activity: 2436

Merit: 6643

be constructive or S.T.F.U

Quote from: seoincorporation on December 18, 2018, 02:55:54 PM

So, the rules were: give a table of data with 3 colums. first the temperature, then the pressure and at last the property we want to get, it could be entropy, density, viscosity and other stuff... Run the ANN ad if we get a R close to 1 (linear regression coefficient), then our equation was right.

in theory you can, but practically you just can't. the difference between the example you gave is that that first both input and output are small, and the result does not have to be 100% accurate, also the output can be reversed with a few possibilities.

for example if you are only allowed to multiply 2 numbers as an input to generate a single output

if you have

3 * 4 = 12

if you have

3 * ? = 12 > easy to solve the second input

if you have

? * ? = 12

you have a few options to try out

1* 12 = 12

2 * 6 = 12

3 * 4 =12

4* 3 = 12
----------------------------------

but doing this with a 256 bits you always have 2^256 combinations !

the strength of the hashing function is that you can't know where it came from.

ABCbits

legendary

Activity: 2870

Merit: 7490

Crypto Swap Exchange

You have better chance to get private key by finding weakness on PRNG/CSPRNG when generate private key/seed or become first people who able to use quantum computer (which is powerful to get private key from public key).

I also remember someone had same thought like you and prepare lots of tools Roll Eyes

seoincorporation

legendary

Activity: 3346

Merit: 3130

Quote from: Foxpup on December 18, 2018, 10:35:30 PM

Quote from: seoincorporation on December 18, 2018, 02:55:54 PM

With Artificial intelligence we can train a neural network to get an equation from a bunch of data.

No, a neural network merely gets a regression from a bunch of correlated data. If the data is uncorrelated, as hashes are, then the resulting regression will be random and have no predictive power whatsoever. This is a basic fact of machine learning. If you think this "in theory is possible", then your "theory" is one that is completely unknown in the machine learning field, and you should perhaps tell us more about how it works on a fundamental level before making grandiose claims about its capabilities.

Thanks for your answer Foxpup, now it's clear why this isn't possible. even with infinite time and energy since the data is uncorrelated isn't possible. And it's sad, i was thinking the ANN was here to solve impossible problems for simple maths but looks like it has limits too.

Foxpup

legendary

Activity: 4542

Merit: 3393

Vile Vixen and Miss Bitcointalk 2021-2023

Quote from: seoincorporation on December 18, 2018, 02:55:54 PM

With Artificial intelligence we can train a neural network to get an equation from a bunch of data.

No, a neural network merely gets a regression from a bunch of correlated data. If the data is uncorrelated, as hashes are, then the resulting regression will be random and have no predictive power whatsoever. This is a basic fact of machine learning. If you think this "in theory is possible", then your "theory" is one that is completely unknown in the machine learning field, and you should perhaps tell us more about how it works on a fundamental level before making grandiose claims about its capabilities.

bitmover

legendary

Activity: 2352

Merit: 6089

bitcoindata.science

Quote from: LoyceV on December 18, 2018, 03:07:05 PM

This won't work, there is no "equation" that returns a one-way hash.

This is specific why Proof of work exists.

To solve the proof of work miners need to scan for a value (called Nonce) that when hashed together with the block data, the hash begins with a number of zeros.

If you could do it backwards, you would just choose a hash and , as you already have Block data, nonce would be easily calculated.

LoyceV

legendary

Activity: 3290

Merit: 16489

Thick-Skinned Gang Leader and Golden Feather 2021

Quote from: seoincorporation on December 18, 2018, 02:55:54 PM

If we give a table with 100 million addys with their privatekeys, AI could be able to get an equation to direct get the private keys from addys.

This won't work, there is no "equation" that returns a one-way hash.
Mycryptopedia explains it better than I can:

Quote from: https://www.mycryptopedia.com/sha-256-related-bitcoin/

Hash functions are powerful because they are ‘one-way’. What this is means is, it is possible for anyone to use a hash function to produce an output when given an input; however, it is impossible to use the output of the hash function to reconstruct its given input.

seoincorporation

legendary

Activity: 3346

Merit: 3130

With Artificial intelligence we can train a neural network to get an equation from a bunch of data. I learn to do this when i was writing my thesis, but what i did was to get an equation to predict any property from water if we had the temp and the pressure.

So, the rules were: give a table of data with 3 colums. first the temperature, then the pressure and at last the property we want to get, it could be entropy, density, viscosity and other stuff... Run the ANN ad if we get a R close to 1 (linear regression coefficient), then our equation was right.

While more information our table have then would be easier to the ANN to get the equation, so, after this introduction here is the question. What if we train a neural network to get private keys from addys? If we give a table with 100 million addys with their privatekeys, AI could be able to get an equation to direct get the private keys from addys. Maybe it would take some years for the ANN to get that equation, but in theory is possible and the same could be done with sha256.

Topic: Artificial neural network VS Bitcoin Address. (Read 289 times)