Adversarial AI
Understanding how to trick Artificial Intelligence.
Operational Phase
01
Model Inversion
Reconstructing private training data from the model.
02
Poisoning Attacks
Corrupting the training data to backlash the AI.
03
Evasion Attacks
Creating adversarial examples (pixel changes) to fool vision models.
04
Model Stealing
Querying a black-box model to replicate it.