Adversarial Reprogramming of Text Classification Neural Networks
This work addresses a security vulnerability in text classification systems, enabling adversaries to repurpose pre-trained models without modification, which is incremental as it extends adversarial reprogramming from continuous to discrete input spaces.
The authors tackled the problem of adversarial reprogramming for text classification neural networks, which have discrete input spaces, by introducing a context-based vocabulary remapping model and training procedures for white-box and black-box settings, successfully repurposing models like LSTM, bi-directional LSTM, and CNN for alternate classification tasks.
Adversarial Reprogramming has demonstrated success in utilizing pre-trained neural network classifiers for alternative classification tasks without modification to the original network. An adversary in such an attack scenario trains an additive contribution to the inputs to repurpose the neural network for the new classification task. While this reprogramming approach works for neural networks with a continuous input space such as that of images, it is not directly applicable to neural networks trained for tasks such as text classification, where the input space is discrete. Repurposing such classification networks would require the attacker to learn an adversarial program that maps inputs from one discrete space to the other. In this work, we introduce a context-based vocabulary remapping model to reprogram neural networks trained on a specific sequence classification task, for a new sequence classification task desired by the adversary. We propose training procedures for this adversarial program in both white-box and black-box settings. We demonstrate the application of our model by adversarially repurposing various text-classification models including LSTM, bi-directional LSTM and CNN for alternate classification tasks.