Qu-ANTI-zation: Exploiting Quantization Artifacts for Achieving Adversarial Outcomes
arXiv:2110.13541
Abstract
Quantization is a popular technique that the parameter representation of a neural network from floating-point numbers into lower-precision ones (, 8-bit integers). It reduces the memory footprint and the computational cost at inference, facilitating the deployment of resource-hungry models. However, the parameter perturbations caused by this transformation result in between the model before and after quantization. For example, a quantized model can misclassify some test-time samples that are otherwise classified correctly. It is not known whether such differences lead to a new security vulnerability. We hypothesize that an adversary may control this disparity to introduce specific behaviors that activate upon quantization. To study this hypothesis, we weaponize quantization-aware training and propose a new training framework to implement adversarial quantization outcomes. Following this framework, we present three attacks we carry out with quantization: (i) an indiscriminate attack for significant accuracy loss; (ii) a targeted attack against specific samples; and (iii) a backdoor attack for controlling the model with an input trigger. We further show that a single compromised model defeats multiple quantization schemes, including robust quantization techniques. Moreover, in a federated learning scenario, we demonstrate that a set of malicious participants who conspire can inject our quantization-activated backdoor. Lastly, we discuss potential counter-measures and show that only re-training consistently removes the attack artifacts. Our code is available at https://github.com/Secure-AI-Systems-Group/Qu-ANTI-zation
arXiv commentAccepted to NeurIPS 2021 [Poster]
References in corpus (10)
- Learning Transferable Visual Models From Natural Language Supervision
- Compressing Deep Convolutional Networks using Vector Quantization
- BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain
- How To Backdoor Federated Learning
- PACT: Parameterized Clipping Activation for Quantized Neural Networks
- Improving Neural Network Quantization without Retraining using Outlier Channel Splitting
- Demon in the Variant: Statistical Analysis of DNNs for Robust Backdoor Contamination Detection
- Can Adversarial Weight Perturbations Inject Neural Backdoors?
- Towards the Limit of Network Quantization
- Breaking Transferability of Adversarial Samples with Randomness