The article, ‘Efficient Trigger Word Insertion’, focuses on the threats posed by backdoor attacks on deep neural network models in the field of Natural Language Processing (NLP). The authors propose a strategy to reduce the number of poisoned samples while maintaining a satisfactory Attack Success Rate (ASR) in text backdoor attacks. They introduce an efficient trigger word insertion strategy involving trigger word optimization and poisoned sample selection. The method is tested on various datasets and models, demonstrating improved attack effectiveness in text classification tasks. The approach achieves an ASR of over 90% with only 10 poisoned samples in the dirty-label setting and requires merely 1.5% of the training data in the clean-label setting.
Publication date: 27 Nov 2023
Project Page: Not Provided
Paper: https://arxiv.org/pdf/2311.13957
