Ensemble Adversarial Defense via Integration of Multiple Dispersed Low Curvature Models

arXiv:2403.16405v1 Announce Type: new
Abstract: The integration of an ensemble of deep learning models has been extensively explored to enhance defense against adversarial attacks. The diversity among sub-models increases the attack cost required to deceive the majority of the ensemble, thereby improving the adversarial robustness. While existing approaches mainly center on increasing diversity in feature representations or dispersion of first-order gradients with respect to input, the limited correlation between these diversity metrics and adversarial robustness constrains the performance of ensemble …

abstract adversarial adversarial attacks arxiv attacks cost cs.cr cs.cv cs.lg deep learning defense diversity ensemble improving integration low multiple robustness type via