Secret AI File Encoding

Secret AI Encoding lets you encode, decode, back up, and share small files using a trained model-based key.

The app creates .aiz files that can be sent through messaging apps, cloud storage, email, or other file-transfer tools. To decode a file, the recipient needs the matching model and, if used, the same mask password.

Secret AI Encoding is designed for users who want an experimental, model-based way to protect small files while keeping full control over their own model files.

Key features:

• Encode and decode files
Select a file, encode it with the app, and create a shareable .aiz output. Decode it later using the matching model.

• Share .aiz files easily
Encoded files can be shared through any messaging service, email app, cloud drive, or file-transfer tool.

• Model-based protection
Files are encoded using a trained key model. For private files, generate, train, secure, or import your own model instead of using the generic default model.

• Optional mask password
Add a mask password for extra protection. The password is mixed with the file salt to generate the payload mask stream. The password itself is not stored inside the .aiz file.

• Model backup and restore
Back up your model, import another one, replace the current model, or create a new model directly from the app. Keep your model safe: files encoded with a lost model may no longer be recoverable.

• Training panel
View training and validation loss while the model is being trained or secured.

• External model generation
For PC-based model generation, use the Python 3.10 project:
https://github.com/DisDoh/Game_Secret_AI

• Small-file workflow
The app is designed for files up to 500 kB because of calculation and device-performance limits. Larger files should be compressed or split before encoding.

How it works:

1. Create, secure, or import a private model.
2. Choose a small file and encode it.
3. Share the generated .aiz file.
4. Share the matching model separately through a trusted channel when possible.
5. The recipient loads the same model, enters the mask password if one was used, and decodes the file.

Important security notes:

Secret AI Encoding is an experimental model-based file protection tool. It is intended to make file recovery impractical without the matching model, but it is not a replacement for professionally audited encryption software.

Security mainly depends on keeping the trained model private. If the model is public, reused too widely, shared together with encoded files, or exposed alongside known original files, recovery attacks may become more realistic.

For best results:

• Generate your own private model.
• Do not use the generic model for sensitive files.
• Keep secure backups of your model.
• Share the model separately from encoded files.
• Do not publish models used for private files.
• Treat an exposed model as compromised.
• Use audited encryption tools for high-risk, regulated, or highly sensitive data.

Limitations:

• Files larger than 500 kB should be compressed or split.
• If the model is lost, encoded files may not be recoverable.
• If someone obtains both the encoded file and the matching model, they may be able to decode the file unless a mask password was used and remains unknown.
• Encoding can fail if the model cannot reconstruct the selected file with full accuracy.

Secret AI Encoding gives you a compact workflow to encode, share, decode, train, import, and back up small files with model-based protection.