Search papers, labs, and topics across Lattice.
This paper introduces SegPAR, a novel class-centric decision-based framework designed to enhance the efficiency of sparse attacks in semantic segmentation. By addressing the limitations of existing methods that suffer from query inefficiency due to pixel-centric accumulation, SegPAR employs a new discrepancy reward mechanism that improves performance in terms of sparsity efficiency and mean Intersection over Union (MIoU) reduction. Extensive experiments demonstrate that SegPAR not only surpasses black-box baselines but also remains competitive with white-box sparse attacks, establishing a new benchmark in this area.
Shifting from pixel-centric to class-centric exploration, SegPAR achieves unprecedented efficiency in sparse attacks for semantic segmentation, outperforming existing methods.
Despite the practical relevance of sparse decision-based black-box threats, they have received limited attention in semantic segmentation. To bridge this gap, we adapt the most representative decision-based black-box sparse attacks from the classification domain to serve as baselines, establishing a rigorous benchmark for this underexplored setting. In this context, we demonstrate that one of the existing methods suffers from severe query inefficiency due to its image-centric pixel accumulation, which rapidly exhausts query budgets across the vast image space. To overcome this, we propose SegPAR, a novel decision-based framework that shifts to a class-centric exploration paradigm. Furthermore, to eliminate the misleading feedback generated by standard decision rewards during pixel accumulation, we introduce a novel discrepancy reward. Extensive experiments show that SegPAR significantly outperforms black-box baselines in sparsity efficiency and MIoU reduction, while remaining competitive with white-box sparse attacks. Code is available at \href{https://github.com/KAU-QuantumAILab/SegPAR}{https://github.com/KAU-QuantumAILab/SegPAR}.
