Abstract
This paper demonstrates a fatal vulnerability in natural language inference (NLI) and text classification systems. More concretely, we present a ‘backdoor poisoning’ attack on NLP models. Our poisoning attack utilizes conditional adversarially regularized autoencoder (CARA) to generate poisoned training samples by poison injection in latent space. Just by adding 1% poisoned data, our experiments show that a victim BERT finetuned classifier’s predictions can be steered to the poison target class with success rates of >80% when the input hypothesis is injected with the poison signature, demonstrating that NLI and text classification systems face a huge security risk.- Anthology ID:
- 2020.findings-emnlp.373
- Volume:
- Findings of the Association for Computational Linguistics: EMNLP 2020
- Month:
- November
- Year:
- 2020
- Address:
- Online
- Editors:
- Trevor Cohn, Yulan He, Yang Liu
- Venue:
- Findings
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 4175–4189
- Language:
- URL:
- https://aclanthology.org/2020.findings-emnlp.373
- DOI:
- 10.18653/v1/2020.findings-emnlp.373
- Cite (ACL):
- Alvin Chan, Yi Tay, Yew-Soon Ong, and Aston Zhang. 2020. Poison Attacks against Text Datasets with Conditional Adversarially Regularized Autoencoder. In Findings of the Association for Computational Linguistics: EMNLP 2020, pages 4175–4189, Online. Association for Computational Linguistics.
- Cite (Informal):
- Poison Attacks against Text Datasets with Conditional Adversarially Regularized Autoencoder (Chan et al., Findings 2020)
- PDF:
- https://preview.aclanthology.org/proper-vol2-ingestion/2020.findings-emnlp.373.pdf
- Code
- alvinchangw/CARA_EMNLP2020 + additional community code
- Data
- MultiNLI, SNLI