Prompt Injection Attacks in LLMs
ResearchSix-category taxonomy of prompt-injection attacks with a heuristic classifier (16 regex patterns) and an 83-test pytest suite. Layered mitigation guidance from input handling through agent isolation.
6
Categories
16
Regex patterns
83
Pytest cases
COVA CCI / Spring 2026 / Python · APA 7