Detecting Contaminated Code-Generation Prompt Batches via Influence Functions

Excerpt

arXiv:2608.14303v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly used for code generation, yet they remain vulnerable to prompts that elicit insecure implementations. Existing defenses typically rely on predefined threat models or known vulnerability patterns, limiting their effectiveness against novel attacks. We propose CodeSIFT, a threat-model-agnostic detection method that leverages influence functions to identify batches of prompts that induce anomalous model be