Summary of Accuracy Of Textfooler Black Box Adversarial Attacks on 01 Loss Sign Activation Neural Network Ensemble, by Yunzhe Xue and Usman Roshan
Accuracy of TextFooler black box adversarial attacks on 01 loss sign activation neural network ensemble
by Yunzhe Xue, Usman Roshan
First submitted to arxiv on: 12 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary A novel study investigates the defense of 01 loss sign activation neural networks against popular black box text adversarial attacks. The researchers test these networks on four text classification datasets: IMDB reviews, Yelp reviews, MR sentiment classification, and AG news classification. They find that their 01 loss sign activation network is significantly more resilient to the TextFooler attack compared to sigmoid activation cross entropy and binary neural networks. A new variation of the 01 loss sign activation convolutional neural network with a novel global pooling step further improves its adversarial accuracy, rendering TextFooler practically useless. The study’s findings suggest that 01 loss sign activation networks could be developed into foolproof models against text adversarial attacks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A team of researchers explored how well 01 loss sign activation neural networks can defend against sneaky text-based attacks. They used four real-life datasets, like movie reviews and news articles, to test their approach. The results show that these special neural networks are much harder to trick than other types. By adding a new step to the network, they made it even more powerful at recognizing fake text. This could lead to creating super-secure models that can’t be fooled by these kinds of attacks. |
Keywords
* Artificial intelligence * Classification * Cross entropy * Neural network * Sigmoid * Text classification
