Recent increases in deploying deep neural networks (DNNs) on resource-constrained devices, combined with the observation that not all input samples require the same amount of computations, have sparked interest in input- adaptive dynamic neural networks (DyNNs). These DyNNs bring in more efficient inferences and enable deploying DNNs on resource-constrained devices e.g. mobile devices. In this work, we study a new vulnerability about DyNNs: can adversaries manipulate a DyNN to provide a false sense of efficiency? To answer this question, we design EfficFrog, an adversarial attack that injects univer- sal efficiency backdoors in DyNNs. EfficFrog poison only a minimal percentage of DyNNs training data to in- ject a backdoor trigger into DyNNs. During the infer- ence time, EfficFrog can slow down backdoored DyNNs and abuse the computational resources of systems running DyNNs by adding the trigger to any inputs - an availability threat analogous to the denial-of-service attacks. We eval- uate EfficFrog on three DNN backbone architectures (based on VGG16, MobileNet, and ResNet56) on two popu- lar datasets (CIFAR-10 and Tiny ImageNet) We show that a EfficFrog reduces the efficiency of DyNNs on triggered input samples while keeping almost the same efficiency on clean samples.