Alexandre Cristovo Maiorano

The paper evaluates prompt-injection defenses for educational LLM tutors, demonstrating that optimal security requires balancing adversarial robustness, usability, and latency, and proposing a comprehensive benchmark protocol for comparison.

The paper measures the specific defensive contribution of various LLM security controls, demonstrating that while defenses like refusal filters and budget controls are effective, they are susceptible to degradation when faced with LLM-generated paraphrasing.

The paper introduces a deterministic method to automatically synthesize initial SIEM detection rules (Sigma rules) from attack simulation findings, ensuring full traceability back to the specific originating test probe.