Non-uniform Illumination Attack for Fooling Convolutional Neural Networks | IEEE Journals & Magazine | IEEE Xplore