DeepSloth: Researchers find denial-of-service equivalent against machine learning systems

Presented at the International Conference on Learning Representations (ICLR), the technique neutralizes optimization techniques that speed up the operation of deep neural networks.