Tom Butcher
8f2d3a2489
- Created .gitignore to exclude build artifacts and dependencies. - Added package.json and package-lock.json for project dependencies and scripts. - Included pnpm workspace configuration for managing packages. - Implemented TypeScript configuration in tsconfig.json. - Added README.md with project description and usage instructions. - Introduced native code for DES encryption and decryption in C/C++. - Created initial decoded data structure for handling scan results. - Established basic file structure for decoded outputs and native builds.
61 lines
2.9 KiB
Markdown
61 lines
2.9 KiB
Markdown
# DES Key Cracker
|
||
|
||
Native desktop app (Node.js + [NodeGui](https://docs.nodegui.org/docs/api/manual/synopsis/)) that recovers the 8-byte DES key for a TechMed / Pectus `.med` scan, using CPU threads plus GPU acceleration:
|
||
|
||
- **macOS:** Apple Metal (`des_brute` kernel)
|
||
- **Windows:** NVIDIA CUDA when the CUDA Toolkit is installed
|
||
- **Both:** multi-core CPU fallback
|
||
|
||
Search is short printable keys over a charset, padded to 8 bytes.
|
||
|
||
Machines can cooperate on a LAN. The **master** listens on a WebSocket port and hands each **slave worker** (each exposed CPU or GPU) a block described only by generation parameters (`start`, `count`, `deviceKey`, `keyLen`, `charset`, `padByte`) — not a list of keys. Each worker has its own block size; CPU workers default to 1,048,576 keys and GPU workers to 16,777,216 so the GPU keeps more of the keyspace.
|
||
|
||
## Requirements
|
||
|
||
- Node.js 18+ (the app runs under **qode**, which NodeGui installs)
|
||
- CMake and a C++ toolchain (needed to install `@nodegui/nodegui`)
|
||
- **macOS:** Xcode command-line tools (clang + Metal)
|
||
- **Windows CPU:** clang++, g++, or MSVC (`cl`)
|
||
- **Windows CUDA:** [CUDA Toolkit](https://developer.nvidia.com/cuda-toolkit) (`nvcc`) and an NVIDIA GPU. Without `nvcc` the app still runs on CPU.
|
||
|
||
## Build
|
||
|
||
```bash
|
||
npm install
|
||
npm run build:native
|
||
npm run build
|
||
npm start
|
||
```
|
||
|
||
`build:native` compiles whatever this machine supports into `bin/`:
|
||
|
||
| Binary | Platform |
|
||
|---|---|
|
||
| `med_cpu` | macOS and Windows |
|
||
| `med_gpu` | macOS Metal |
|
||
| `med_cuda` | Windows (or `FORCE_CUDA=1` if `nvcc` is on the PATH) |
|
||
|
||
## Using the app
|
||
|
||
1. Set **Role** in the toolbar to Master or Slave.
|
||
2. Under **Devices**, tick local cards to use or expose. On the master, set **Block size** on each local and remote card; **Workers** (CPU thread count) appears only on CPU cards.
|
||
3. **Master:** choose the `.med` file, charset, min/max length, and pads. Optionally **Start service** so other computers can join. Click **Start search**. The master searches on its selected local devices and assigns a block of that device’s size to every exposed slave device.
|
||
4. **Slave:** tick the devices to expose, set CPU threads on CPU cards if needed, enter the master’s IP and port, and **Connect**. The slave does not set block size — the master does. Each exposed device stays listed on **Workers** with current-block progress and a completed-block count. The slave does not need the `.med` file.
|
||
|
||
Hits are written under `decoded/bruteforce/` (`decrypted_*.bin`, `hit_*.json`, `results.json`).
|
||
|
||
Do not expose the WebSocket port to the public internet; it has no authentication.
|
||
|
||
## Native runner CLI
|
||
|
||
All three runners share the same commands:
|
||
|
||
```
|
||
med_* devices
|
||
med_* selftest [--device N]
|
||
med_* brute --device N --key-len L --charset S --pad XX \
|
||
--target HEX --fills HEX,... --start I --count N --batch B
|
||
```
|
||
|
||
Metal also needs `--shader native/metal/des_bruteforce.metal`. CPU accepts `--workers N`.
|