GradMDM: Adversarial Attack on Dynamic Networks

Rahmani, Hossein (2023) GradMDM: Adversarial Attack on Dynamic Networks. IEEE Transactions on Pattern Analysis and Machine Intelligence. ISSN 0162-8828 (In Press)

Text (Layer_wise_Attack)
Layer_wise_Attack.pdf - Accepted Version
Restricted to Repository staff only until 1 January 2050.
Available under License Creative Commons Attribution.
Download (881kB)

Abstract

Dynamic neural networks can greatly reduce computation redundancy without compromising accuracy by adapting their structures based on the input. In this paper, we explore the robustness of dynamic neural networks against \textit{energy-oriented attacks} targeted at reducing their efficiency. Specifically, we attack dynamic models with our novel algorithm GradMDM. GradMDM is a technique that adjusts the direction and the magnitude of the gradients to effectively find a small perturbation for each input, that will activate more computational units of dynamic models during inference. We evaluate GradMDM on multiple datasets and dynamic models, where it outperforms previous energy-oriented attack techniques, significantly increasing computation complexity while reducing the perceptibility of the perturbations.