PRALab

MLSec Laboratory

Minimum-norm attack ensembles for query-controlled, curve-based adversarial robustness evaluation.

SecML-Torch: A Library for Robustness Evaluation of Deep Learning Models

Implementation for "Counterfeit Answers: Adversarial Forgery against OCR-Free Document Visual Question Answering"

A security-aware framework for evaluating machine learning anomaly detection in 5G Core networks against realistic adversarial threats and evasion attacks.