Syntax-Oriented Shortcut: A Syntax Level Perturbing Algorithm for Preventing Text Data From Being Learned

Summary

Researchers propose UTE-SS, an algorithm that makes text examples unlearnable by adding syntax-level perturbations that models learn as a shortcut. The method uses a syntax template generator to select an optimal syntax per category and a perturbing text generator to rewrite in-class texts with that template. Experiments on eight Transformer-based pretrained language models across four NLP tasks show its effectiveness and flexibility, and the code is public at https://github.com/libolb/UTE-SS.

Related items

• MediumCVE-2026-121936: easyVoice server-side request forgery through openaiBaseUrl in /api/v1/ttsSimilar attack · NVD/CVE Database
• MediumCVE-2026-118628: MLflow get-artifacts server-side request forgery information disclosureSimilar attack · NVD/CVE Database
• HighCVE-2026-118625: MLflow Tracking Server privilege escalation via get_endpointsSimilar attack · NVD/CVE Database
• InfoWhen trust becomes the payload: Inside the threats, etc quarterly reportSimilar attack · Elastic Security Labs
• MediumCritical Langflow flaws enable authentication bypass and RCESimilar attack · CSO Online