Adversarial Black-Box Attacks On Text Classifiers Using Multi-Objective Genetic Optimization Guided By Deep Networks
This addresses security vulnerabilities in text classification systems, though it is incremental as it builds on existing adversarial attack methods.
The paper tackles the problem of generating black-box adversarial examples to fool neural network text classifiers, achieving an average attack success rate of 65.67% on SST and 36.45% on IMDB datasets, with improvements of 49.48% and 101% over DeepWordBug.
We propose a novel genetic-algorithm technique that generates black-box adversarial examples which successfully fool neural network based text classifiers. We perform a genetic search with multi-objective optimization guided by deep learning based inferences and Seq2Seq mutation to generate semantically similar but imperceptible adversaries. We compare our approach with DeepWordBug (DWB) on SST and IMDB sentiment datasets by attacking three trained models viz. char-LSTM, word-LSTM and elmo-LSTM. On an average, we achieve an attack success rate of 65.67% for SST and 36.45% for IMDB across the three models showing an improvement of 49.48% and 101% respectively. Furthermore, our qualitative study indicates that 94% of the time, the users were not able to distinguish between an original and adversarial sample.