Secure LLM Gateway: Defending AI from Prompt Injection & Data Leaks

Engineered by Mehran Rasool — Full-Stack Developer

A five-stage hybrid security gateway for LLM applications defending against direct prompt injection, jailbreaking, and sensitive PII leaks across multiple languages.

Key Results (Verified 150-Prompt Evaluation)

• Overall Accuracy: 82.7%
• PII Recall: 100%
• Block Precision: 100%
• Average Latency: 525ms