Search for: All records | NSF Public Access

Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher. Some full text articles may not yet be available without a charge during the embargo (administrative interval).
What is a DOI Number?

Some links on this page may take you to non-federal websites. Their policies may differ from this site.

Analyzing and Defending against Membership Inference Attacks in Natural Language Processing Classification

https://doi.org/10.1109/BigData55660.2022.10020711

Wang, Yijue ; Xu, Nuo ; Huang, Shaoyi ; Mahmood, Kaleel ; Guo, Dan ; Ding, Caiwen ; Wen, Wujie ; Rajasekaran, Sanguthevar ( December 2022 , IEEE Big Data)

Full Text Available
Position: TrustLLM: Trustworthiness in Large Language Models

Huang, Yue ; Sun, Lichao ; Wang, Haoran ; Wu, Siyuan ; Zhang, Qihui ; Li, Yuan ; Gao, Chujie ; Huang, Yixin ; Lyu, Wenhan ; Zhang, Yixuan ; et al ( July 2024 , PMLR)

Free, publicly-accessible full text available July 29, 2025
Against Membership Inference Attack: Pruning is All You Need

https://doi.org/10.24963/ijcai.2021/432

Wang, Yijue ; Wang, Chenghong ; Wang, Zigeng ; Zhou, Shanglin ; Liu, Hang ; Bi, Jinbo ; Ding, Caiwen ; Rajasekaran, Sanguthevar ( August 2021 , International Joint Conference on Artificial Intelligence)

The large model size, high computational operations, and vulnerability against membership inference attack (MIA) have impeded deep learning or deep neural networks (DNNs) popularity, especially on mobile devices. To address the challenge, we envision that the weight pruning technique will help DNNs against MIA while reducing model storage and computational operation. In this work, we propose a pruning algorithm, and we show that the proposed algorithm can find a subnetwork that can prevent privacy leakage from MIA and achieves competitive accuracy with the original DNNs. We also verify our theoretical insights with experiments. Our experimental results illustrate that the attack accuracy using model compression is up to 13.6% and 10% lower than that of the baseline and Min-Max game, accordingly.

more » « less
Full Text Available
TAG: Gradient Attack on Transformer-based Language Models

https://doi.org/10.18653/v1/2021.findings-emnlp.305

Deng, Jieren ; Wang, Yijue ; Li, Ji ; Wang, Chenghong ; Shang, Chao ; Liu, Hang ; Rajasekaran, Sanguthevar ; Ding, Caiwen ( January 2021 , The 2021 Conference on Empirical Methods in Natural Language Processing)

Full Text Available
A Secure and Efficient Federated Learning Framework for NLP

https://doi.org/10.18653/v1/2021.emnlp-main.606

Wang, Chenghong ; Deng, Jieren ; Meng, Xianrui ; Wang, Yijue ; Li, Ji ; Lin, Sheng ; Han, Shuo ; Miao, Fei ; Rajasekaran, Sanguthevar ; Ding, Caiwen ( January 2021 , EMNLP)

Full Text Available