Mind Your Heart: Stealthy Backdoor Attack on Dynamic Deep Neural Network in Edge Computing

TL;DR

This paper introduces a stealthy backdoor attack targeting dynamic multi-exit deep neural networks in edge computing, which remains hidden until the model is transformed into its multi-exit form, evading detection.

Contribution

It presents a novel backdoor injection method on vanilla DNNs that activates only in their dynamic multi-exit architectures, demonstrating effectiveness and stealthiness.

Findings

Effective attack on multiple DNN architectures and datasets.

Backdoor remains undetected by current detection methods.

Stealthy activation only after model transformation.

Abstract

Transforming off-the-shelf deep neural network (DNN) models into dynamic multi-exit architectures can achieve inference and transmission efficiency by fragmenting and distributing a large DNN model in edge computing scenarios (e.g., edge devices and cloud servers). In this paper, we propose a novel backdoor attack specifically on the dynamic multi-exit DNN models. Particularly, we inject a backdoor by poisoning one DNN model's shallow hidden layers targeting not this vanilla DNN model but only its dynamically deployed multi-exit architectures. Our backdoored vanilla model behaves normally on performance and cannot be activated even with the correct trigger. However, the backdoor will be activated when the victims acquire this model and transform it into a dynamic multi-exit architecture at their deployment. We conduct extensive experiments to prove the effectiveness of our attack on three structures (ResNet-56, VGG-16, and MobileNet) with four datasets (CIFAR-10, SVHN, GTSRB, and Tiny-ImageNet) and our backdoor is stealthy to evade multiple state-of-the-art backdoor detection or removal methods.