Coasties concerned about online safety may soon see advancements in detecting hidden toxic messages, thanks to innovative research led by Dr Johnny Chan, a lecturer at the University of Auckland.

His newly developed tool addresses a growing challenge in online communication—users masking harmful words to bypass moderation filters.

Common tactics include replacing letters with numbers or symbols, such as “Y0u’re st00pid,” or embedding harmful words within compound phrases like “IdiotFace.”

Some even add spaces or extra characters to obscure toxic terms, effectively slipping past traditional keyword-based moderation systems.

Dr Chan’s team has introduced a pre-processing technique designed to enhance moderation tools’ ability to identify these disguised messages.

“Our work ensures the existing systems function as effectively as they should, even when faced with disguised toxic messages,” the team noted in a statement.

The tool works in three key stages.

First, it simplifies text by removing unnecessary elements, such as excessive punctuation or extraneous characters.

Second, it standardises misspellings, phrasing, and grammar to uncover harmful intent.

Finally, it identifies patterns in disguised words, such as “h@te” or “I d i o t,” and restructures them for easier detection.

Dr Johnny Chan.

By addressing these tactics, the tool enhances content moderation across platforms, safeguarding users—especially vulnerable groups—from harmful online behaviour.

It holds promise for a variety of applications, including social media platforms, businesses, and organisations monitoring public discourse.

Dr Chan emphasised the broader potential: “This tool helps protect younger audiences, counter negative brand campaigns, and maintain respectful public dialogue. It marks an important step in creating safer, more inclusive online environments.”

While the research focuses on tackling hidden toxicity in text, Dr Chan’s team acknowledges this is only the beginning.

Future advancements may explore deeper contextual analysis, paving the way for even more effective moderation in dynamic and diverse online spaces.