Zero-Knowledge Attack for Replicating Protected Deep Neural Networks | IEEE Conference Publication | IEEE Xplore