2022-06-13 更新
Adversarial RAW: Image-Scaling Attack Against Imaging Pipeline
Authors:Junjian Li, Honglong Chen
Deep learning technologies have become the backbone for the development of computer vision. With further explorations, deep neural networks have been found vulnerable to well-designed adversarial attacks. Most of the vision devices are equipped with image signal processing (ISP) pipeline to implement RAW-to-RGB transformations and embedded into data preprocessing module for efficient image processing. Actually, ISP pipeline can introduce adversarial behaviors to post-capture images while data preprocessing may destroy attack patterns. However, none of the existing adversarial attacks takes into account the impacts of both ISP pipeline and data preprocessing. In this paper, we develop an image-scaling attack targeting on ISP pipeline, where the crafted adversarial RAW can be transformed into attack image that presents entirely different appearance once being scaled to a specific-size image. We first consider the gradient-available ISP pipeline, i.e., the gradient information can be directly used in the generation process of adversarial RAW to launch the attack. To make the adversarial attack more applicable, we further consider the gradient-unavailable ISP pipeline, in which a proxy model that well learns the RAW-to-RGB transformations is proposed as the gradient oracles. Extensive experiments show that the proposed adversarial attacks can craft adversarial RAW data against the target ISP pipelines with high attack rates.
PDF
论文截图
Plug & Play Attacks: Towards Robust and Flexible Model Inversion Attacks
Authors:Lukas Struppek, Dominik Hintersdorf, Antonio De Almeida Correia, Antonia Adler, Kristian Kersting
Model inversion attacks (MIAs) aim to create synthetic images that reflect the class-wise characteristics from a target classifier’s private training data by exploiting the model’s learned knowledge. Previous research has developed generative MIAs that use generative adversarial networks (GANs) as image priors tailored to a specific target model. This makes the attacks time- and resource-consuming, inflexible, and susceptible to distributional shifts between datasets. To overcome these drawbacks, we present Plug & Play Attacks, which relax the dependency between the target model and image prior, and enable the use of a single GAN to attack a wide range of targets, requiring only minor adjustments to the attack. Moreover, we show that powerful MIAs are possible even with publicly available pre-trained GANs and under strong distributional shifts, for which previous approaches fail to produce meaningful results. Our extensive evaluation confirms the improved robustness and flexibility of Plug & Play Attacks and their ability to create high-quality images revealing sensitive class characteristics.
PDF Accepted by ICML 2022
论文截图