Bitcoin puzzle transaction ~32 BTC prize to who solves it

[nochkin]

hopeless:
Scanning 23,298,085,122,481 Electrum v1 seed combinations using 8 cores...
[5.0s] Checked: 5,108,500 | Speed: 1,021,699/s | Progress: 0.000022%

It does not look like your code is actually doing its job. It does not even try to search for the address, this is why the speed is for doing basically nothing.

you did not even try.

Actually, I did and even tried to debug it. There are a few issues with it, this is why it does not search.

[citb0in]

does anyone know a fork or similar program like keyhunt-cuda that works for p2sh addresses (addresses with prefix 3........) ?

[kTimesG]

does anyone know a fork or similar program like keyhunt-cuda that works for p2sh addresses (addresses with prefix 3........) ?

And break what? There's an infinite amount of possible redemption scripts. There's nothing that says you'd find a matching script hash for a single script entry, you would then have to try all permutations for two entries, and then three, and so on. There might also be other script ops thrown into the mix, all of them need to be tried as well, until some script hash matches.

[tmar777]

If I understood your question right, maybe one possible way to simplify the check is to first verify whether both points are n-torsion points (nP = O and nQ = O). If not, you can already rule out that they belong to the same n-order subgroup. If they both are, i think you have to build up the entire subgroup as you said, i don't think there is a direct way to check. But at least the n-torsion check can help "quickly" exclude cases that definitely don’t match, or at least use less computation. Could this make sense?

I used the wrong terminology. I meant quick-checking whether two points belong to the same orbit. Since n - 1 divides by 2, 3, 149, 631, .... So all points on the curve can be grouped in orbits of size 2, 3, 149, 631. The sum of all points (and scalars) of each distinct orbit is O (0).

In the end, any valid point has a unique 7-dimensional scalar. I think a trial and error search of common coefficients on lower dimensions (same orbit, or finding the distance between orbits) may simplify the ECDLP (with some sort of pollard rho). It might also be nothin'.

Can you elaborate more on this? Send me a DM. I am working on this idea.

[kTimesG]

In the end, any valid point has a unique 7-dimensional scalar. I think a trial and error search of common coefficients on lower dimensions (same orbit, or finding the distance between orbits) may simplify the ECDLP (with some sort of pollard rho). It might also be nothin'.

Can you elaborate more on this? Send me a DM. I am working on this idea.

Let's drill through the ECC torus hyper-dimensions and crack open the orbitals faster than a MIT graduate can re-engineer a cat's laser toy to emit 256-bit math in a 16x16 dots grid light in real-time!

The trick is easy: let's say we have some valid k as a private key. It's gonna have this form because of the n - 1 upper bound:

k = c1 * a + c2 * b + c3 * c + c4 *d + c5 * e + c6 * f + c7

c1 to c7 are the unknown ranks of each "digit". Every digit has an upper bound (product of the lower-level  rank factors).

Mission? Find k somehow, given all the properties we can take advantage of (endomorphisms for the 3-orbital, symmetry for the 2-orbital, lambda products and zero-sum of orbit points... you get the point).

Build it up as a lattice or some sort of HNP. Use heuristics to search for k + delta to optimize your equations system (e.g. move around the public key, maybe in a really smart way!).

I ain't got time for this though. Some screws didn't fit back anymore, and my cat is angry that I broke its laser again. BTW never try to solder 256 diodes by hand.

You can mention me in the thank you notes inside the first page's footer of your breakthrough paper, though.

[citb0in]

does anyone know a fork or similar program like keyhunt-cuda that works for p2sh addresses (addresses with prefix 3........) ?

And break what? There's an infinite amount of possible redemption scripts. There's nothing that says you'd find a matching script hash for a single script entry, you would then have to try all permutations for two entries, and then three, and so on. There might also be other script ops thrown into the mix, all of them need to be tried as well, until some script hash matches.

sorry, forgot to mention that. It's not related to puzzles. Just playin' around

[nochkin]

does anyone know a fork or similar program like keyhunt-cuda that works for p2sh addresses (addresses with prefix 3........) ?

And break what? There's an infinite amount of possible redemption scripts. There's nothing that says you'd find a matching script hash for a single script entry, you would then have to try all permutations for two entries, and then three, and so on. There might also be other script ops thrown into the mix, all of them need to be tried as well, until some script hash matches.

sorry, forgot to mention that. It's not related to puzzles. Just playin' around

For playing around you could ask ChatGPT to make something in Python.

[Cringineer]

Ive always had an idea for a new type of mining/key pool.

The idea I had was a lottery type pool. Ie, if 5 people are mining, and 1 has the majority of power. What if instead of them winning the entire reward/sharing it with others, the person who finds the key gets 50-60-70-80% of the prize, then the remaining % is distributed equally to all parties?

It may not be a great idea, but for people like me who have really bad hardware, gives people like me and them a chance to get something for our troubles.

I assume people with high hash/key rates would probably avoid such a pool, but it gives people with less powerful hardware a place to consolidate their efforts.


Really curious to know what all your thoughts on this are.

[teguh54321]

No new method ? 😅

Im running 12x3080ti ... vram 100-106c  Today 2 card dead after running kangoroo for 3 month 😔🙃

Hope break 135 before all dead 🤪

[MrGPBit]

No new method ? 😅

Im running 12x3080ti ... vram 100-106c  Today 2 card dead after running kangoroo for 3 month 😔🙃

Hope break 135 before all dead 🤪

No good cooling

[analyticnomad]

Hey guys, can anyone point me in the right direction for C++ with gpu optimization and cuda integration/implementation?

Something I can learn from and try to absorb? I've been constantly researching and watching Youtube vids but they're more geared towards gaming and deep learning AI models. I just can't seem to pull together the coding logic or process to be successful. The way I think about everything is still so abstract at this point.

This is such i niche thing we're doing and accessing the right info from the right people is key.

For context (and for people to laugh at me, which I can't blame you) I just bought a 5090, AMD 9950x specifically to learn a new OS, learn C++ and to attempt solving these puzzles.

Knowing exactly zero about any of this a few months ago, to creating python scripts with ChatGPT (this is where everyone laughs) I have learned a lot but have reached the peak of what my little Macbook Air can do.

I've looked at Fixed Paul, JLP and a few other guys on Github and tbh, it might as well be hieroglyphics, but one small step at a time, I'm forcing myself to learn just a little every day. I'm also paranoid downloading the wrong .exe will F-start my life.

Any solid info, ideas and harsh criticism is welcomed.  

[fecell]

Hey guys, can anyone point me in the right direction for C++

of course!

Programming Massively Parallel Processor

https://safari.ethz.ch/architecture/fall2019/lib/exe/fetch.php?media=2013_programming_massively_parallel_processors_a_hands-on_approach_2nd.pdf

[tmar777]

In the end, any valid point has a unique 7-dimensional scalar. I think a trial and error search of common coefficients on lower dimensions (same orbit, or finding the distance between orbits) may simplify the ECDLP (with some sort of pollard rho). It might also be nothin'.

Can you elaborate more on this? Send me a DM. I am working on this idea.

Let's drill through the ECC torus hyper-dimensions and crack open the orbitals faster than a MIT graduate can re-engineer a cat's laser toy to emit 256-bit math in a 16x16 dots grid light in real-time!

The trick is easy: let's say we have some valid k as a private key. It's gonna have this form because of the n - 1 upper bound:

k = c1 * a + c2 * b + c3 * c + c4 *d + c5 * e + c6 * f + c7

c1 to c7 are the unknown ranks of each "digit". Every digit has an upper bound (product of the lower-level  rank factors).

Mission? Find k somehow, given all the properties we can take advantage of (endomorphisms for the 3-orbital, symmetry for the 2-orbital, lambda products and zero-sum of orbit points... you get the point).

Build it up as a lattice or some sort of HNP. Use heuristics to search for k + delta to optimize your equations system (e.g. move around the public key, maybe in a really smart way!).

I ain't got time for this though. Some screws didn't fit back anymore, and my cat is angry that I broke its laser again. BTW never try to solder 256 diodes by hand.

You can mention me in the thank you notes inside the first page's footer of your breakthrough paper, though.

How can this be done without any side-channel info?
What degree do you hold?

[kTimesG]

Hey guys, can anyone point me in the right direction for C++ with gpu optimization and cuda integration/implementation?

C++ won't help you, the GPU doesn't care about any fancy object oriented programming. Besides, C++ isn't like a magical unicorn that solves everything and is a perfect language, anyone who lost more than 20 years learning it will definitely tell you they are still learning it. If they don't, they are lying or are unaware that there's always something new to learn about it. Which makes them really bad developers.

To learn what you want, it all really depends on your background. Try this:

1. Run a Hello world program that prints from the GPU.
2. Decompose everything that happened before delving into hacking advanced cryptographic problems:

a. Understand what the host code did (and draw the line between your C++ or whatever code and the GPU code)
b. Understand what the GPU did - dump the CUBIN, check the SASS ops, learn about the registers, shared memory, constant memory, basically read the manual.
c. Upgrade your hello world program into a different programming language (maybe Python) but load and run the exact same kernel as the first time. Notice how it has nothing to do with any C++ this time around, since you're loading a CUBIN into the GPU.

d. Now you're a pro - apply for lead tech jobs at emerging de-fi startups. Just kidding. You're on your way to fight with compiler bugs, unexpected behaviours, unslept nights scrolling through NVidia forums, and profiling performance of a dozen variations of algorithms to see which one runs better.

Good luck.

[analyticnomad]

Hey guys, can anyone point me in the right direction for C++ with gpu optimization and cuda integration/implementation?

C++ won't help you, the GPU doesn't care about any fancy object oriented programming. Besides, C++ isn't like a magical unicorn that solves everything and is a perfect language, anyone who lost more than 20 years learning it will definitely tell you they are still learning it. If they don't, they are lying or are unaware that there's always something new to learn about it. Which makes them really bad developers.

To learn what you want, it all really depends on your background. Try this:

1. Run a Hello world program that prints from the GPU.
2. Decompose everything that happened before delving into hacking advanced cryptographic problems:

a. Understand what the host code did (and draw the line between your C++ or whatever code and the GPU code)
b. Understand what the GPU did - dump the CUBIN, check the SASS ops, learn about the registers, shared memory, constant memory, basically read the manual.
c. Upgrade your hello world program into a different programming language (maybe Python) but load and run the exact same kernel as the first time. Notice how it has nothing to do with any C++ this time around, since you're loading a CUBIN into the GPU.

d. Now you're a pro - apply for lead tech jobs at emerging de-fi startups. Just kidding. You're on your way to fight with compiler bugs, unexpected behaviours, unslept nights scrolling through NVidia forums, and profiling performance of a dozen variations of algorithms to see which one runs better.

Good luck.

Even this is super advanced for me but I will take this opportunity and brute-force implement tonight. Thank you for the suggestion.

You mentioned "really depends on your background". Did you mean coding background? If so, absolutely none. I have a medical background, specifically paramedic. A dead person is also known as a "code". So the only "coding" I ever did was on dead people, and ironically, this project may be the death of me.

[kTimesG]

You mentioned "really depends on your background". Did you mean coding background? If so, absolutely none. I have a medical background, specifically paramedic. A dead person is also known as a "code". So the only "coding" I ever did was on dead people, and ironically, this project may be the death of me.

Oh. No worries, this puzzle's been analyzed to death by everyone, from toddlers to NSA counter-operatives hiding in Siberian igloos, ice-fishing between GitHub pushes.

Medical background can help. After all, you just made me realize that "dead code" is a redundant pleonasm [sigh]. But specifically for elliptic curves, we do have some limbs to worry about - do we do 4x64 or 5x52? Careful though - going too deep into the guts of this thing without knowing what your fingers are doing can severely cause injuries. Memory issues can often lead to a straight-up zombie process that needs to be killed. But then an exorcist is needed, to fix any existing daemon leftovers, or you'll be left wondering why the heck the issue repeats itself after a clean compile.

Why not start off with some autopsy of existing codes before getting your hands dirty though? Though I feel sorry that even an experienced coder would scratch his head for weeks, trying to read through the gibberish of the really ugly style used in most public projects, before giving up and hitting the "Create new directory" button.

[analyticnomad]

You mentioned "really depends on your background". Did you mean coding background? If so, absolutely none. I have a medical background, specifically paramedic. A dead person is also known as a "code". So the only "coding" I ever did was on dead people, and ironically, this project may be the death of me.

Oh. No worries, this puzzle's been analyzed to death by everyone, from toddlers to NSA counter-operatives hiding in Siberian igloos, ice-fishing between GitHub pushes.

Medical background can help. After all, you just made me realize that "dead code" is a redundant pleonasm [sigh]. But specifically for elliptic curves, we do have some limbs to worry about - do we do 4x64 or 5x52? Careful though - going too deep into the guts of this thing without knowing what your fingers are doing can severely cause injuries. Memory issues can often lead to a straight-up zombie process that needs to be killed. But then an exorcist is needed, to fix any existing daemon leftovers, or you'll be left wondering why the heck the issue repeats itself after a clean compile.

Why not start off with some autopsy of existing codes before getting your hands dirty though? Though I feel sorry that even an experienced coder would scratch his head for weeks, trying to read through the gibberish of the really ugly style used in most public projects, before giving up and hitting the "Create new directory" button.

Bridging the two together. Clever. Poetry. Love it.

[kTimesG]

Nothing is changing. The same shit-stream from the main-branch bullshit-master.

Hello. You have a critical multi-threading bug in your PointsSearch repository.

Code:

bf.insert(someHexString)

Ignoring the fact that you are effectively using twice as much RAM then needed, the more problematic issue is that you are totally missing any form of race condition checks, resulting in potential corruption of your bf object internal state, which is not thread-safe.

So yeah, nothing is changing indeed. Script kiddies wanna-be coders f**ing it up as they go along.

[jeLACOSTE]

Guys, have you ever thought that maybe it’s not a piece-by-piece puzzle at all, but simply a quiz?
I mean, maybe the creator hid a pattern that could lead us straight to the solution. Has this idea been discussed already?
Could you give us some hints, please @saatoshi_rising?
We would really appreciate it!

I am the creator.

You are quite right, 161-256 are silly.  I honestly just did not think of this.  What is especially embarrassing, is this did not occur to me once, in two years.  By way of excuse, I was not really thinking much about the puzzle at all.

I will make up for two years of stupidity.  I will spend from 161-256 to the unsolved parts, as you suggest.  In addition, I intend to add further funds.  My aim is to boost the density by a factor of 10, from 0.001*length(key) to 0.01*length(key).  Probably in the next few weeks.  At any rate, when I next have an extended period of quiet and calm, to construct the new transaction carefully.

A few words about the puzzle.  There is no pattern.  It is just consecutive keys from a deterministic wallet (masked with leading 000...0001 to set difficulty).  It is simply a crude measuring instrument, of the cracking strength of the community.

Finally, I wish to express appreciation of the efforts of all developers of new cracking tools and technology.  The "large bitcoin collider" is especially innovative and interesting!

[kTimesG]

I have tested it in  a step by step manner. On my machine everything works fine with point_search.
And moreover it is just a concept code. As almost everything that can be found in the open-source domain.
What is the purpose of your presence here?

That's pretty much how MT bugs present themselves: Heisenbugs that eventually manifest sooner or later. So your code is conceptually broken, even if it passes your tests. Also, are you aware that debugging a MT program is pretty much useless when running concurrent threads, especially if you have breakpoints that are reached one at a time? So of course everything works fine in that scenario, since debugging a concurrent function is basically identical as if running it single-threaded. Tests are not even needed to spot it as a bug. This is simple CS 101. At some point, at some time, you'll have two threads accessing the same bloom filter RAM, and the CPU cache lines will screw up the bloom filter because of R/W lack of sync.

Well, my presence here is definitely not to show off broken code to n00bz that think it actually works correctly, while calling other people that know their thing as main-branch BS-ers. See it as you wish, and take from it what you want.