All You Need is "Love": Evading Hate-speech Detection
This work highlights critical vulnerabilities in hate-speech detection systems, which is an incremental but important issue for social media platforms and content moderation.
The paper tackled the problem of hate-speech detection by showing that state-of-the-art models perform poorly when tested on different data types and are vulnerable to adversarial attacks like typos or word changes, with character-level features proving more resistant than word-level ones.
With the spread of social networks and their unfortunate use for hate speech, automatic detection of the latter has become a pressing problem. In this paper, we reproduce seven state-of-the-art hate speech detection models from prior work, and show that they perform well only when tested on the same type of data they were trained on. Based on these results, we argue that for successful hate speech detection, model architecture is less important than the type of data and labeling criteria. We further show that all proposed detection techniques are brittle against adversaries who can (automatically) insert typos, change word boundaries or add innocuous words to the original hate speech. A combination of these methods is also effective against Google Perspective -- a cutting-edge solution from industry. Our experiments demonstrate that adversarial training does not completely mitigate the attacks, and using character-level features makes the models systematically more attack-resistant than using word-level features.