TextAttack

An open-source Python framework for adversarial attacks and robustness evaluation of natural language processing models. It supports multiple text adversarial attack strategies and is widely used in security testing of NLP models to assess their vulnerability to input perturbations and to improve model robustness.