一种基于粒子群优化的对抗攻击补丁生成方法
By improving the particle swarm optimization algorithm and fitness function design, adversarial patches with strong offensive capabilities and natural appearance in the physical world are generated. This solves the problems of high computational resource consumption and insufficient concealment in existing methods, and achieves efficient and covert adversarial attack effects.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING INST OF TECH
- Filing Date
- 2025-09-15
- Publication Date
- 2026-07-17
AI Technical Summary
Existing adversarial patch generation methods suffer from problems such as high computational resource consumption, difficulty in effective conversion in the physical world, lack of concealment in generated patches, and difficulty in balancing attack effects with naturalness.
An improved particle swarm optimization algorithm is used to search for adversarial patches in a multidimensional parameter space. Combined with fitness function design, adversarial patches with strong attack power and natural appearance are generated. The position, shape and color of the patches are optimized through a multi-patch collaborative strategy.
It improves patch generation efficiency, ensures effectiveness and stealth in the physical world, increases attack success rate, and addresses the shortcomings of existing methods in terms of computational resources and stealth.
Smart Images
Figure CN121457501B_ABST