Residue-Based Natural Language Adversarial Attack Detection

Anonymous

Paper Link: https://openreview.net/forum?id=eFGgjI4Wk-V

Paper Type: Long paper (up to eight pages of content + unlimited references and appendices)

Abstract: Deep learning based systems are susceptible to adversarial attacks, where a small, imperceptible change at the input alters the model prediction. However, to date the majority of the approaches to detect these attacks have been designed for image processing systems. Many popular image adversarial detection approaches are able to identify adversarial examples from embedding feature spaces, whilst in the NLP domain existing state of the art detection approaches solely focus on input text features, without consideration of model embedding spaces. This work examines what differences result when porting these image designed strategies to Natural Language Processing (NLP) tasks - these detectors are found to not port over well. This is expected as NLP systems have a very different form of input: discrete and sequential in nature, rather than the continuous and fixed size inputs for images. As an equivalent model-focused NLP detection approach, this work proposes a simple sentence-embedding "residue" based detector to identify adversarial examples. On many tasks, it out-performs ported image domain detectors and recent state of the art NLP specific detectors.

Copyright Consent Signature (type Name Or NA If Not Transferrable): Vyas Raina

Copyright Consent Name And Address: University of Cambridge Department of Engineering, Trumpington St, Cambridge, CB2 1PZ, UK.