Tweeted By @ml_review
TextBugger: Generating Adversarial Text Against Real-world Applications #NDSS2019
— ML Review (@ml_review) December 19, 2018
1. Outperforms SoTA in attack success rate
2. Preserves the utility of benign text, ~95% of text human comprehendible
3. Comp. complexity sub-linear to the text lengthhttps://t.co/ihiR3g32wk pic.twitter.com/xPnzXBKRp7