2020 · 40 citations · 27 references
ComputingComputer ScienceCybersecurityComputer SecurityMachine LearningNatural Language ProcessingDeep LearningInjection VulnerabilitiesLanguage ModelingRepresentation LearningAutomatic Text ClassificationBackdoor TechniquesInput HypothesisLatent Space ExplorationNatural Language InferenceNatural Language InterfacesSecurity Risk ManagementText Classification
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.
27
Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks
Jun-Yan Zhu, Taesung Park, Phillip Isola et al. · 2017 · 21.9K citations · Full text
Conditional Generative Adversarial Nets
Mehdi Mirza, Simon Osindero · arXiv (Cornell University) · 2014 · 8.9K citations · Full text
Intriguing properties of neural networks
Christian Szegedy, Wojciech Zaremba, Ilya Sutskever et al. · arXiv (Cornell University) · 2013 · 5.7K citations · Full text