opencl_vanity_gpg

使用 GPU（OpenCL）快速生成带有“靓号”的 PGP 密钥！

“靓号”指的是带有连号等特定格式的密钥指纹或 ID（例如以 77777777 结尾），具体介绍和生成原理请参见：

• 一位 PGP 进步青年的科学算号实践
• 某科学的 PGP 算号指南

简单来说，密钥指纹是密钥生效时间和公钥内容的 SHA-1，通过不断生成密钥和修改时间（成本更低）的暴力遍历方式找到“靓号”。

Tip

姊妹项目：使用 WebGL 实现的网页版 TransparentLC/webgl-vanity-gpg

实际上这个项目也可以当成是它的“锈化”版…_φ(･ω･` )

目前最好的同类工具是使用 CPU 的 RedL0tus/VanityGPG 和使用 GPU（CUDA）的 cuihaoleo/gpg-fingerprint-filter-gpu。这个项目是使用 GPU（OpenCL）实现的，优点当然是开箱即用了。

GitHub Actions 有自动编译好的二进制文件。

使用方法

$ opencl_vanity_gpg -h
Usage: opencl_vanity_gpg [OPTIONS]

Options:
  -c, --cipher-suite <CIPHER_SUITE>  Cipher suite of the vanity key
                                     ed25519, ecdsa-****, rsa**** => Primary key
                                     cv25519,  ecdh-****          => Subkey
                                     Use gpg CLI for further editing of the key. [default: ed25519] [possible values: ed25519, cv25519, rsa2048, rsa3072, rsa4096, ecdh-p256, ecdh-p384, ecdh-p521, ecdsa-p256, ecdsa-p384, ecdsa-p521]
  -u, --user-id <USER_ID>            OpenPGP compatible user ID [default: "Dummy <[email protected]>"]
  -p, --pattern <PATTERN>            A pattern less than 40 chars for matching fingerprints
                                     > Format:
                                     * 0-9A-F are fixed, G-Z are wildcards
                                     * Other chars will be ignored
                                     * Case insensitive
                                     > Example:
                                     * 11XXXX** may output a fingerprint ends with 11222234 or 11AAAABF
                                     * 11XXYYZZ may output a fingerprint ends with 11223344 or 11AABBCC
  -f, --filter <FILTER>              OpenCL kernel function for uint h[5] for matching fingerprints
                                     Ignore the pattern and no estimate is given if this has been set
                                     > Example:
                                     * (h[4] & 0xFFFF)     == 0x1234     outputs a fingerprint ends with 1234
                                     * (h[0] & 0xFFFF0000) == 0xABCD0000 outputs a fingerprint starts with ABCD
  -o, --output <OUTPUT>              The dir where the vanity keys are saved
  -d, --device <DEVICE>              Device ID to use
  -t, --thread <THREAD>              Adjust it to maximum your device's usage
  -i, --iteration <ITERATION>        Adjust it to maximum your device's usage [default: 512]
      --timeout <TIMEOUT>            Exit after a specified time in seconds
      --oneshot                      Exit after getting a vanity key
      --no-progress                  Don't print progress
      --no-secret-key-logging        Don't print armored secret key
      --list-device                  Show available OpenCL devices then exit
  -h, --help                         Print help
  -V, --version                      Print version

$ opencl_vanity_gpg -p 11XXYYZZ --oneshot
[2025-01-08T19:00:25Z INFO  opencl_vanity_gpg] Using device: Apple M1 Pro
[2025-01-08T19:00:25Z INFO  opencl_vanity_gpg] Auto set thread: 1048576
[2025-01-08T19:00:25Z INFO  opencl_vanity_gpg] You will get vanity keys created after 2008-01-05T00:11:53.021Z
[2025-01-08T19:00:25Z WARN  opencl_vanity_gpg] No output dir given. Generated vanity keys will not be saved.
[2025-01-08T19:00:26Z INFO  opencl_vanity_gpg::utils::vanity_key] Get a vanity key:
-----BEGIN PGP PRIVATE KEY BLOCK-----
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==
=ZeYm
-----END PGP PRIVATE KEY BLOCK-----

[2025-01-08T19:00:26Z INFO  opencl_vanity_gpg::utils::vanity_key] Created at: 2008-01-06T19:40:20.000Z (1199648420)
[2025-01-08T19:00:26Z INFO  opencl_vanity_gpg::utils::vanity_key] Fingerprint #0: B9C8FAB0EEA22949DB183A71E45E037111CC6622
[2025-01-08T19:00:26Z INFO  opencl_vanity_gpg::utils::vanity_key] Fingerprint #1: 7FBC6D063D9C4CF4C281E8A14876C8831552E9B4
[2025-01-08T19:00:26Z INFO  opencl_vanity_gpg] Hashed: 536.87M (512.00x) Time: 1.53s Speed: 351.76M hash/s

一个 Curve25519 的密钥由用来签名和认证的 Ed25519 主密钥和用来加密的 Cv25519 子密钥组成。VanityGPG 只能生成主密钥为“靓号”的密钥，而这个项目也可以生成子密钥为“靓号”的密钥（只需要添加参数 -c cv25519 即可）。对于其他 NISP P-*** 的椭圆曲线的密钥也是类似的。

性能对比

Repo	计算方式	速度（hash/s）	注释
RedL0tus/VanityGPG	CPU	165m	Xeon w5-2465X ×16 cores Arch Linux
这个项目	CPU	120m	Xeon w5-2465X ×16 cores Arch Linux CPU 也可以是 OpenCL 的计算设备，虽然性能损耗比较严重……
cuihaoleo/gpg-fingerprint-filter-gpu	GPU	1b	A16 1/8 Ubuntu 24.04
这个项目	GPU	1.5b	A16 1/8 Ubuntu 24.04
TransparentLC/webgl-vanity-gpg	GPU	2b	GTX 1070 Windows 11
这个项目	GPU	3b	GTX 1070 Windows 11
TransparentLC/webgl-vanity-gpg	GPU	7b	RTX A5500 Windows 11
这个项目	GPU	12b	RTX A5500 Windows 11
这个项目	GPU	380m	Apple M1 Pro
这个项目	GPU	1.6b	Apple M4 Pro

以上的速度均为生成 Curve25519 类型的密钥的速度。