Attacks and Defenses for Free-Riders in Multi-Discriminator GAN
Not on the map. Why?
Abstract (arXiv)
Generative Adversarial Networks (GANs) are increasingly adopted by the industry to synthesize realistic images. Due to data not being centrally available, Multi-Discriminator (MD)-GANs training framework employs multiple discriminators that have direct access to the real data. Distributedly training a joint GAN model entails the risk of free-riders, i.e., participants that aim to benefit from the common model while only pretending to participate in the training process. In this paper, we conduct the first characterization study of the impact of free-riders on MD-GAN. Based on two production prototypes of MD-GAN, we find that free-riders drastically reduce the ability of MD-GANs to produce images that are indistinguishable from real data, i.e., they increase the FID score -- the standard measure to assess the quality of generated images. To mitigate the model degradation, we propose a defense strategy against free-riders in MD-GAN, termed DFG. DFG distinguishes free-riders and benign participants through periodic probing and clustering of discriminators' responses based on a reference response of free-riders, which then allows the generator to exclude the detected free-riders from the training. Furthermore, we extend our defense, termed DFG+, to enable discriminators to filter out free-riders at the variant of MD-GAN that allows peer exchanges of discriminators networks. Extensive evaluation on various scenarios of free-riders, MD-GAN architecture, and three datasets show that our defenses effectively detect free-riders. With 1 to 5 free-riders, DFG and DFG+ averagely decreases FID by 5.22% to 11.53% for CIFAR10 and 5.79% to 13.22% for CIFAR100 in comparison to an attack without defense. In a shell, the proposed DFG(+) can effectively defend against free-riders without affecting benign clients at a negligible computation overhead.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 21 of its 40 references that are in the map. First posted on arXiv 24 Jan 2022.
Builds on
21 of 40 references.
References · 21
- 2016Communication-Efficient Learning of Deep Networks from Decentralized Datacited by 7,592
- 2015Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classificationcited by 3,590
- 2018Large Scale GAN Training for High Fidelity Natural Image Synthesiscited by 2,609
- 2017Improved Training of Wasserstein GANscited by 2,223
- 2020Evaluation: from precision, recall and F-measure to ROC, informedness, markedness and correlationcited by 420
- 2019Federated Machine Learning: Concept and Applicationscited by 409
- 2021CTAB-GAN: Effective Table Data Synthesizingcited by 156
- 2020FedGAN: Federated Generative Adversarial Networks for Distributed Datacited by 89
- 2016Generative Multi-Adversarial Networkscited by 60
- 2020Free-rider Attacks on Model Aggregation in Federated Learningcited by 57
- 2018MD-GAN: Multi-Discriminator Generative Adversarial Networks for Distributed Datasetscited by 49
- 2019Free-riders in Federated Learning: Attacks and Defensescited by 48
9 more
- 2020Synthetic Learning: Learn From Distributed Asynchronized Discriminator GAN Without Sharing Medical Image Datacited by 38
- 2017Recurrent Topic-Transition GAN for Visual Paragraph Generationcited by 25
- 2018Voice Impersonation using Generative Adversarial Networkscited by 16
- 2021Fed-TGAN: Federated Learning Framework for Synthesizing Tabular Datacited by 9
- 2020Learn distributed GAN with Temporary Discriminatorscited by 5
- 2021ComicGAN: Text-to-Comic Generative Adversarial Networkcited by 4
- 2019ZK-GanDef: A GAN based Zero Knowledge Adversarial Training Defense for Neural Networkscited by 3
- 2017Distributional Adversarial Networkscited by 2
- 2021Enhancing Robustness of On-line Learning Models on Highly Noisy Datacited by 2
Built on by
Citing papers per year: 2022: 1.
1 paper.