Launching today
MIRAGE
AI honeypot that deceives attackers instead of blocking them
2 followers
AI honeypot that deceives attackers instead of blocking them
2 followers
Most AI security tools silently block attacks. MIRAGE does the opposite — when it detects a prompt injection, jailbreak, or data exfiltration attempt, it routes the attacker to a convincing decoy persona that feeds them fabricated data and honey tokens. Nothing real leaks. Every move is logged and tagged with MITRE ATLAS techniques for threat intelligence. Built with Go + Redis + Gemini/OpenAI. Docker Compose, no build step. Demo mode works without any API keys.
