Hello!
I see that there are two separate modes for Ethereum:
15600 = Ethereum Wallet, PBKDF2-HMAC-SHA256
15700 = Ethereum Wallet, PBKDF2-SCRYPT
When looking in my keystore file I see the cypher is aes-128-ctr and "scrypt" is referenced as the kdf:
{
"crypto" : {
"cipher" : "aes-128-ctr",
"cipherparams" : {
"iv" : "..."
},
"ciphertext" : "...",
"kdf" : "scrypt",
"kdfparams" : {
"dklen" : 32,
"n" : 262144,
"r" : 1,
"p" : 8,
"salt" : "..."
},
"mac": "...",
"id": "...",
"version": 3
}
}
Which mode should I be using?
Use 15700. You probably want to use -D1 after installing CPU driver with latest hashcat beta version.
Awesome, thanks. I have CUDA installed but I believe I'm using the device manufacturer's driver. Is there a special driver needed?
I'm running NVIDIA 2070.
Thanks for the help.
scrypt is GPU resistant . You'd be running 15700 mode on CPU only.
@atom what's added in latest beta about this?
I see. It's all coming back to me, I tried to reclaim my wallet a few years ago and using the CPU sounds like what I was doing.
My processor looks to be running about 17/hs which is quite slow for the amount of guesses it has to go through. Is it possible to use a Raspberry Pi B or Zero to run Hashcat, is that supported for 15700 using CPU?
scrypt is slow, as long as you dont have a very good clue for your password (pw + rules) bruteforcing will just be a waste of time (till the next big bang
)
@CATuGHTI
There are multiple important small changes to the scrypt-based kernel code for instance these:
-
https://github.com/hashcat/hashcat/commi...5f851e9f25
-
https://github.com/hashcat/hashcat/commi...5667a3ca11
-
https://github.com/hashcat/hashcat/commi...4313ade8c1
etc
In theory they should make scrypt perform (much) better, especially on modern hardware etc. These are of course implementation-specific changes that are only of interst to devs and of course users that use/crack scrypt a lot, the user doesn't really need to care too much about these changes except that they should notice a speed improvement in the beta and upcoming release versions (but this of course also depends a lot on the scrypt parameters N, r, p etc and on the users hardware... for very heavy scrypt settings of course the speed improvement can't be noticed that much, because Nrp will make them kind of negligible). The improvement is still huge in general and very important.
@while1
no, you need the opposite, instead of scaling down, you need to scale up (or even better: first think A LOT about the feasibility and keyspace, number of password candidates, patterns etc etc etc). Instead of trying to use a Raspberry PI, that has (compared to a modern CPU or GPU rig, but we already stated here that for scrypt a CPU-rig might be better) almost no computing power, you need to scale up and use modern performant hardware that has a lot of computing power. hashcat doesn't support Raspberry PI for several obvious reasons (almost no computing power, no fan/cooling, not suitable for most of the advanced hard cracking jobs etc). In case of a scrypt-based algorithm you are better of with some modern Intel i9 processors or modern high-performant Intel Xeon processors (or similar). Good luck cracking.
(05-14-2021, 08:19 AM)philsmd Wrote: [ -> ]@while1
no, you need the opposite, instead of scaling down, you need to scale up (or even better: first think A LOT about the feasibility and keyspace, number of password candidates, patterns etc etc etc). Instead of trying to use a Raspberry PI, that has (compared to a modern CPU or GPU rig, but we already stated here that for scrypt a CPU-rig might be better) almost no computing power, you need to scale up and use modern performant hardware that has a lot of computing power. hashcat doesn't support Raspberry PI for several obvious reasons (almost no computing power, no fan/cooling, not suitable for most of the advanced hard cracking jobs etc). In case of a scrypt-based algorithm you are better of with some modern Intel i9 processors or modern high-performant Intel Xeon processors (or similar). Good luck cracking.
Thanks, my total guesses are around ~400-500million. Unfortunately, Hashcat (for good reason) doesn't support Raspberry Pi. I benchmarked a Pyhton library making attempts on a Raspberry Pi B+ and am getting 2.11 attempts a second. I think I'm going to group a dozen together in a beowulf cluster and have them run along side my computer when that's available to work on it. Should take a few months
If I want to scale up where would I find a CPU-Rig or directions to build??
well every motherboard over 2 cpu sockets is more or less not really "consumer hardware" and used inside serverblades or similar
you could try to workaround this with multiple machines and using hashtopolis to distribute the work to these machines