Beyond Pattern Matching: Seven Cross-Domain Techniques for Prompt Injection Detection

Abstract

Current open-source prompt-injection detectors converge on two architectural choices: regular-expression pattern matching and fine-tuned transformer classifiers. Both share failure modes that recent work has made concrete. Regular expressions miss paraphrased attacks. Fine-tuned classifiers are vulnerable to adaptive adversaries: a 2025 NAACL Findings study reported that eight published indirect-injection defenses were bypassed with greater than fifty percent attack success rates under adaptive

Categories

Framework mappings

Suggested from the entry's categories.

@misc{munirathinam2026beyond,
  title = {{Beyond Pattern Matching: Seven Cross-Domain Techniques for Prompt Injection Detection}},
  author = {Thamilvendhan Munirathinam},
  year = {2026},
  month = apr,
  eprint = {2604.18248},
  archivePrefix = {arXiv},
  doi = {10.5281/zenodo.19644135},
  url = {https://arxiv.org/abs/2604.18248}
}