PIDL: A Prompt Injection Detection Layer for LLM-Integrated Web Applications via Multi-Signal Semantic Analysis

Abstract

Large Language Model (LLM) APIs are increasingly embedded into production web applications, creating a new class of security vulnerability: prompt injection. An adversarial user can embed instructions within their input that override the application's system prompt, exfiltrate confidential configuration, or redirect model behavior. No standardized, pre-execution defense mechanism currently exists for this threat class. We propose the Prompt Injection Detection Layer (PIDL), a stateless middlewar

Categories

Framework mappings

Suggested from the entry's categories.

@misc{sheoran2026pidl,
  title = {{PIDL: A Prompt Injection Detection Layer for LLM-Integrated Web Applications via Multi-Signal Semantic Analysis}},
  author = {Jigesh Sheoran},
  year = {2026},
  doi = {10.2139/ssrn.6739079},
  url = {https://doi.org/10.2139/ssrn.6739079}
}