This is a multithreaded implementation of Polynonce, the polynomial nonce recurrence attack on ECDSA.

In this repository, you will find the following.

In the attacks directory: multithreaded implementation of the attacks for Bitcoin, Ethereum and TLS, to be used with inputs produced by ecdsa-dump-bitcoin, ecdsa-dump-ethereum and ecdsa-dump-tls respectively. Note that the dump files must have been grouped by public key, and then, within each group, sorted by timestamp so that the attack works. This is explained in detail further below.

In the original-attack folder, we provide the original (easy to read) proof-of-concept of the attack. The one that was later used in production is the recurrence_nonce attack.

In the results folder, we provide the results we obtained for the Half-Half attack.

• Sagemath 9.6+

First, install sagemath for your platform. For example, on Arch Linux:

sudo pacman -S sagemath

Then install the required python libraries within the sage environment:

sage -pip install -r requirements.txt

cd original-attack/
./recurrence_nonces.py

First obtain a dump using ecdsa-dump-bitcoin. The input file must have the following format:

r;s;pubkey;txid;message_hash;block_time

Then sort the dump by public key and then by timestamp (sort by field 3 and then 6):

sort bitcoin-input.csv --parallel=16 -T /path/to/tmp/ --field-separator ';' --buffer-size="70%" -k3,3 -k6,6 > sorted-bitcoin-input.csv

Finally, run the attack on the sorted file:

./attacks/ecdsa_bitcoin_attack.py -i sorted-bitcoin-input.csv -o bitcoin-attack-results.csv

The output file will contain, on each line (assuming N=4):

d;pubkey;first_txid;batch_size;pubkey_address;r1;r2;r3;r4;s1;s2;s3;s4;hash1;hash2;hash3;hash4;k1;k2;k3;k4;block_time1;block_time2;block_time3;block_time4

Where:

• d is the private key
• r1 to rN are the ECDSA signature r values for each signature in the window where the attack worked
• s1 to sN are the ECDSA signature s values for each signature in the window where the attack worked
• k1 to kN are the recovered ECDSA nonces for each signature in the window where the attack worked
• The other values' names are self-explanatory

From the previous dump of transaction, it is also possible to run our implementation of the Half-Half attack (without lattice attack).

$ python half_parallel_attack.py  sorted-bitcoin-input.csv -b 500000 -o half_half_results.csv

The output file will contain, on each line:

pubkey;privkey;txid;k1;k2

Where k1 and k2 are the two nonces which allowed to recover the private key.

First obtain a dump using ecdsa-dump-ethereum. The input file must have the following format:

from_address;r;s;pubkey;txid;message_hash;block_time

Then sort the dump by public key and then by timestamp (sort by field 4 and then 7):

time sort ethereum-input.csv --parallel=16 -T /path/to/tmp/ --field-separator ';' --buffer-size="70%" -k4,4 -k7,7 > sorted-ethereum-input.csv

Finally, run the attack on the sorted file:

./attacks/ecdsa_ethereum_attack.py -i sorted-ethereum-input.csv -o ethereum-attack-results.csv

The output file will contain, on each line:

d;pubkey;first_txid;batch_size;pubkey_address;r1;r2;r3;r4;s1;s2;s3;s4;hash1;hash2;hash3;hash4;k1;k2;k3;k4;block_time1;block_time2;block_time3;block_time4

Where:

• d is the private key
• r1 to rN are the ECDSA signature r values for each signature in the window where the attack worked
• s1 to sN are the ECDSA signature s values for each signature in the window where the attack worked
• k1 to kN are the recovered ECDSA nonces for each signature in the window where the attack worked
• The other values' names are self-explanatory

First obtain a dump using ecdsa-dump-tls. The input file must have the following format:

r;s;signature_value_hex;pubkey_hex;src_addr;server_name;msg_hex;timestamp

Then sort the dump by public key and then by timestamp (sort by field 4 and then 8):

time sort tls-input.csv --parallel=16 -T /path/to/tmp/ --field-separator ';' --buffer-size="70%" -k4,4 -k8,8 > sorted-tls-input.csv

Finally, run the attack on the sorted file:

./attacks/ecdsa_tls_attack.py -i sorted-tls-input.csv -o tls-attack-results.csv

The output file will contain, on each line (assuming N=4):

d;pubkey;server_name;batch_size;first_ip_address;r1;r2;r3;r4;s1;s2;s3;s4;hash1;hash2;hash3;hash4;k1;k2;k3;k4;timestamp1;timestamp2;timestamp3;timestamp4

Where:

• d is the private key
• r1 to rN are the ECDSA signature r values for each signature in the window where the attack worked
• s1 to sN are the ECDSA signature s values for each signature in the window where the attack worked
• k1 to kN are the recovered ECDSA nonces for each signature in the window where the attack worked
• The other values' names are self-explanatory

To display the full help with all the options available, pass --help or -h:

$ ./attacks/ecdsa_tls_attack.py --help
usage: ecdsa-tls-attack [-h] --input INPUT_PATH --output OUTPUT_PATH [-n N] [--max-futures MAX_FUTURES] [--no-sliding-window]

Run the polynomial nonce recurrence attack on a TLS dataset.

options:
  -h, --help            show this help message and exit
  --input INPUT_PATH, -i INPUT_PATH
                        Path to the input file. Must be a dump file in the format produced by ecdsa-dump-tls and which has been sorted by public key and then by timestamp.
  --output OUTPUT_PATH, -o OUTPUT_PATH
                        Path to the output file to dump to
  -n N                  Number of signatures per batch. N must be >= 4
  --max-futures MAX_FUTURES
                        Maximum number of futures to process in a batch. Increase this number if more cores are available.
  --no-sliding-window   Do not use a sliding window.Only use the first N signatures of each pubkey and discard the rest.Note that this will run faster but some vulnerable signatures may remain undetected

Special thanks to my colleague Marco Macchetti for the original proof-of-concept code in the original-attack directory, for coming up with this idea and for his feedback.

Copyright(c) 2023 Nagravision SA.

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License version 3 as published by the Free Software Foundation.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/.