tool arXiv Sep 15, 2025 · Sep 2025
Asim Waheed, Vasisht Duddu, Rui Zhang et al. · University of Waterloo · Zhejiang University +1 more
Open-source Python library revealing unintended cross-risk tradeoffs when combining ML defenses against adversarial, privacy, and fairness threats
Input Manipulation Attack Membership Inference Attack visiontabular
Machine learning (ML) models are susceptible to various risks to security, privacy, and fairness. Most defenses are designed to protect against each risk individually (intended interactions) but can inadvertently affect susceptibility to other unrelated risks (unintended interactions). We introduce Amulet, the first Python library for evaluating both intended and unintended interactions among ML defenses and risks. Amulet is comprehensive by including representative attacks, defenses, and metrics; extensible to new modules due to its modular design; consistent with a user-friendly API template for inputs and outputs; and applicable for evaluating novel interactions. By satisfying all four properties, Amulet offers a unified foundation for studying how defenses interact, enabling the first systematic evaluation of unintended interactions across multiple risks.
cnn transformer traditional_ml University of Waterloo · Zhejiang University · Aalto University
ML Security Papers