Skip to content

Topic

Model extraction attacks

Techniques that recover proprietary details of a closed machine learning model, from architecture dimensions and output-layer weights to full parameter copies, using only permitted queries against it.

Current clusters