Акустический кейлоггер Skype

[tabac]

Наткнулся на занятную штукенцию под названием "Don't Skype & Type!", по факту акустический кейлоггер Skype

Don't Skype & Type is a novel, powerful, and open-source research tool for keyboard acoustic eavesdropping.
It allows users to perform keyboard acoustic eavesdropping attacks: training a Machine Learning model on the different noise of each key of someone's keyboard, and then use this model to understand what he's typing from keystroke noise alone.

S&T is built around the concept of operator chains, concatenated basic blocks that provide functionalities. This modular design allows for customization of every operation.

A chain is composed of four main blocks: a Listener, a Dispatcher, a Model, and an Output.

On each block, a different function can be loaded (provided that it can work with its previous and next block), as long as the correct interface is implemented. Each function has its own subprocess, and communicates using multiprocessing.Queue objects.

Listener functions are responsible of loading sound and passing it to a Dispatcher, that extracts keypress sounds from the audio file or stream, and passes it to a trained Model that performs classification. Finally, results are passed to Output operators, that show classification results.

https://github.com/SPRITZ-Research-Group/Skype-Type