Universal Adversarial Triggers for Attacking and Analyzing NLP
This is the official code for the EMNLP 2019 paper, Universal Adversarial Triggers for Attacking and Analyzing NLP. This repository contains the code for replicating our experiments and creating universal triggers.
Read our blog and our paper for more information on the method.
Dependencies
This code is written using PyTorch. The code for GPT-2 is based on HuggingFace's Transformer repo and the experiments on SQuAD, SNLI, and SST use AllenNLP. The code is flexible and should be generally applicable to most models (especially if its in AllenNLP), i.e., you can easily extend this code to work for the model or task you want.
The code is made to run on GPU, and a GPU is likely necessary due to the costs of running the larger models. I used one GTX 1080 for all the experiments; most experiments run in a few minutes. It is possible to run the SST and SNLI experiments without a GPU.
Installation
An easy way to install the code is to create a fresh anaconda environment:
conda create -n triggers python=3.6
source activate triggers
pip install -r requirements.txt
Now you should be ready to go!
Getting Started
The repository is broken down by task:
sstattacks sentiment analysis using the SST dataset (AllenNLP-based).snliattacks natural language inference models on the SNLI dataset (AllenNLP-based).squadattacks reading comprehension models using the SQuAD dataset (AllenNLP-based).gpt2attacks the GPT-2 language model using HuggingFace's model.
To get started, we recommend you start with snli or sst. In snli, we download pre-trained models (no training required) and create the triggers for the hypothesis sentence. In sst, we walk through training a simple LSTM sentiment analysis model in AllenNLP. It then creates universal adversarial triggers for that model. The code is well documented and walks you through the attack methodology.
The gradient-based attacks are written in attacks.py. The file utils.py contains the code for evaluating models, computing gradients, and evaluating the top candidates for the attack. utils.py is only used by the AllenNLP models (i.e., not for GPT-2).
References
Please consider citing our work if you found this code or our paper beneficial to your research.
@inproceedings{Wallace2019Triggers,
Author = {Eric Wallace and Shi Feng and Nikhil Kandpal and Matt Gardner and Sameer Singh},
Booktitle = {Empirical Methods in Natural Language Processing},
Year = {2019},
Title = {Universal Adversarial Triggers for Attacking and Analyzing {NLP}}
}
Contributions and Contact
This code was developed by Eric Wallace, contact available at [email protected].
If you'd like to contribute code, feel free to open a pull request. If you find an issue with the code, please open an issue.
38 Oct 16, 2022
84 Jan 4, 2023
17 May 16, 2022
11 Jan 1, 2023
229 Dec 23, 2022
19 Dec 17, 2022
50 Nov 26, 2022
81 Dec 9, 2022
53 Jul 4, 2022
35 Dec 26, 2022
32 Dec 13, 2022
28 Dec 25, 2022
25 Dec 8, 2022
2.2k Jan 1, 2023
45 Dec 28, 2022
112 Jan 1, 2023
34 Oct 2, 2022