White-Box Attacks on Hate-speech BERT Classifiers in German with Explicit and Implicit Character Level Defense

TL;DR

This paper assesses the robustness of German Hate Speech BERT classifiers against novel white-box character and word-level attacks, and compares two new character-level defense strategies to enhance model security.

Contribution

It introduces two novel white-box attack methods at character and word levels and compares two new character-level defenses for German Hate Speech BERT classifiers.

Findings

New attack methods effectively challenge existing defenses.

Character-level defenses show varying robustness against attacks.

Evaluation highlights strengths and weaknesses of defense strategies.

Abstract

In this work, we evaluate the adversarial robustness of BERT models trained on German Hate Speech datasets. We also complement our evaluation with two novel white-box character and word level attacks thereby contributing to the range of attacks available. Furthermore, we also perform a comparison of two novel character-level defense strategies and evaluate their robustness with one another.