Institution profile

Key Laboratory of Cyberspace Security

Academic institutionasia · cn
Research library1linked papers
Opportunities0open roles
Selected work

Representative Papers

Sequential Difference Maximization: Generating Adversarial Examples via Multi-Stage Optimization

Aug 31, 2025

To address the low efficiency of adversarial example generation in robustness evaluation of computer vision models, this paper proposes a multi-stage gradient optimization framework. Methodologically, it introduces (1) a directional probability difference ratio (DPDR) loss function that explicitly maximizes the margin between the upper bound of non-target class probabilities and the true-class probability; (2) a hierarchical “cyclic–stage–step” optimization architecture that progressively suppresses irrelevant class responses to incrementally strengthen attack potency; and (3) joint optimization of the negative true-class probability and DPDR loss to balance attack success rate and computational efficiency. Extensive experiments on ImageNet and other benchmarks demonstrate that the method surpasses state-of-the-art approaches in attack success rate, query efficiency, and cross-model transferability. Moreover, it significantly enhances the defensive performance of adversarial training.

0 citationsRead paper
Recent publications

Latest Papers

Sequential Difference Maximization: Generating Adversarial Examples via Multi-Stage Optimization

Aug 31, 2025

To address the low efficiency of adversarial example generation in robustness evaluation of computer vision models, this paper proposes a multi-stage gradient optimization framework. Methodologically, it introduces (1) a directional probability difference ratio (DPDR) loss function that explicitly maximizes the margin between the upper bound of non-target class probabilities and the true-class probability; (2) a hierarchical “cyclic–stage–step” optimization architecture that progressively suppresses irrelevant class responses to incrementally strengthen attack potency; and (3) joint optimization of the negative true-class probability and DPDR loss to balance attack success rate and computational efficiency. Extensive experiments on ImageNet and other benchmarks demonstrate that the method surpasses state-of-the-art approaches in attack success rate, query efficiency, and cross-model transferability. Moreover, it significantly enhances the defensive performance of adversarial training.

0 citationsRead paper