Attacks on NLP

Multi-objective Attack

Adversarial attack seeking simultaneously to deceive the model while optimizing multiple constraints such as readability or semantic preservation.

← Indietro