A high-performance, CPU-optimized tool for computing public keys on the secp256k1 elliptic curve, with features for searching compressed & uncompressed public keys and customizable search parameters.
- π Fixed 256-bit modular arithmetic
- π Group inversion for point addition operation
- π Precomputed table for points multiplication
- π Search for compressed & uncompressed public keys (hash160)
- π Utilizes SIMD for optimized sha256 (uses SHA extensions, both ARM and Intel)
- π Works seamlessly on MacOS and Linux
- π§ Customizable search range and thread count for flexible usage
git clone https://github.com/vladkens/ecloop.git && cd ecloop
make build
Note: Build has been tested with clang
. It may work with gcc14
, but this has not been thoroughly tested. If anyone knows how to get this to work with gcc14 or earlier β I'd be happy to get a PR.
Usage: ecloop <cmd> [-t <threads>] [-f <filepath>] [-a <addr_type>] [-r <range>]
Compute commands:
add - search in given range with batch addition
mul - search hex encoded private keys (from stdin)
Compute options:
-f <file> - filter file to search (list of hashes or bloom fitler)
-o <file> - output file to write found keys (default: stdout)
-t <threads> - number of threads to run (default: 1)
-a <addr_type> - address type to search: c - addr33, u - addr65 (default: c)
-r <range> - search range in hex format (example: 8000:ffff, default all)
-q - quiet mode (no output to stdout; -o required)
Other commands:
blf-gen - create bloom filter from list of hex-encoded hash160
bench - run benchmark of internal functions
bench-gtable - run benchmark of ecc multiplication (with different table size)
-f
is filter file with hash160 to search. Can be list of hex encoded hashes (one per line) or bloom fitler (must have .blf
extension). -t
use 4 threads. r
β start:end of search range. -o
file where found keys should be saved (if not provided stdout
fill be used). No -a
option provided, so c
(compressed) hash160 will be checked.
ecloop add -f data/btc-puzzles-hash -t 4 -r 800000:ffffff -o /tmp/found.txt
cat privkeys.txt
β source of HEX encoded priv keys to search (can be file or generator program). -f
β hash160 to search as bloom filter (can have false positive results, but has a much smaller size; eg. all BTC addresses ever used have size ~6GB). -a
β what type of hash160 to search (c
β compressed, u
β uncopressed, cu
check both). -t
use 8 threads.
cat privkeys.txt | ecloop mul -f data/btc-puzzles.blf -a cu -t 4
ecloop
can also take a raw word list and automatically hash it with sha256. Use -raw
flag to it.
cat wordlist.txt | ecloop mul -f data/btc-puzzles.blf -a cu -t 4 -raw
cat
reads the list of hex-encoded hash160 values from a file. -n
specifies the number of entries for the Bloom filter (count of hashes). -o
defines the output where to write filter (.blf
extension requried).
Bloom filter uses p = 0.000001 (1 in 1,000,000 false positive). You can adjusting this option by playing with n
. See Bloom Filter Calculator. List of all addressed can be found here.
cat data/btc-puzzles-hash | ecloop blf-gen -n 1024 -o /tmp/test.blf
Then created bloom filter can be used in ecloop
as filter:
ecloop add -f /tmp/test.blf -t 4 -r 8000:ffffff
Note: Bloom filter works with both add
and mul
commands.
Get performance of different function for single thread:
ecloop bench
Should print output like:
_ec_jacobi_add1: 6.52M it/s ~ 0.92s
_ec_jacobi_add2: 5.26M it/s ~ 1.14s
_ec_jacobi_dbl1: 5.42M it/s ~ 1.11s
_ec_jacobi_dbl2: 7.57M it/s ~ 0.79s
ec_jacobi_mul: 0.02M it/s ~ 0.57s
ec_gtable_mul: 0.29M it/s ~ 1.73s
ec_affine_add: 0.30M it/s ~ 1.67s
ec_affine_dbl: 0.30M it/s ~ 1.69s
_fe_modinv_binpow: 0.20M it/s ~ 0.51s
_fe_modinv_addchn: 0.31M it/s ~ 0.32s
addr33: 4.95M it/s ~ 1.01s
addr65: 4.41M it/s ~ 1.14s
This project is written to learn the math over elliptic curves in cryptocurrencies. Functionality as a search for Bitcoin Puzzles is added as a real-world use case.