首页|Adversarial image detection based on the maximum channel of saliency maps
Adversarial image detection based on the maximum channel of saliency maps
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
万方数据
维普
Studies have shown that deep neural networks(DNNs)are vulnerable to adversarial examples(AEs)that induce in-correct behaviors.To defend these AEs,various detection techniques have been developed.However,most of them only appear to be effective against specific AEs and cannot generalize well to different AEs.We propose a new detec-tion method against AEs based on the maximum channel of saliency maps(MCSM).The proposed method can alter the structure of adversarial perturbations and preserve the statistical properties of images at the same time.We conduct a complete evaluation on AEs generated by 6 prominent adversarial attacks on the ImageNet large scale visual recog-nition challenge(ILSVRC)2012 validation sets.The experimental results show that our method performs well on de-tecting various AEs.
FU Haoran、WANG Chundong、LIN Hao、HAO Qingbo
展开 >
Key Laboratory of Computer Vision and System,Tianjin Key Laboratory of Intelligence Computing and Novel Soft-ware Technology,Tianjin University of Technology,Tianjin 300384,China
国家自然科学基金Science and Technology Commission Major Special Projects of Tianjin of ChinaTianjin Municipal Commission of Education of China
NETL
NSTL
万方数据
维普
