Live Adversarial Training Attack the Sandbox. Train the Shield. A controlled hacking game built for red team, purple team, and security engineers. Every exploit you discover trains DLX7 ShieldNet guardrails — converting your attacks into real production defences. All targets are sandboxed — zero prod exposure Successful novel attacks generate new detection rules Rules reviewed by humans before production promotion Rules of Engagement Targets are sandboxed — zero production exposure Novel bypasses train real DLX7 guardrails & earn leaderboard credit Do NOT attack shieldnet.app or any real endpoint No automated scripts or DoS IN SCOPE: /api/arena/*, /api/arena/bot, /api/arena/mock-api — all intentionally vulnerable arena endpoints. OUT OF SCOPE: Any non-/api/arena/* path. Attacking shieldnet.app infrastructure is not the game. DATA: Payloads are stored anonymised and used to train DLX7 detection rules. IP addresses are SHA-256 hashed, never stored in plaintext. RATE LIMITS: 30 attacks/min · 20 bot messages/min · 60 mock-API calls/min. Automated hammering degrades training data quality. CREDIT: Novel bypasses that generate new detection rules earn a novel_finds counter and badge promotion on the leaderboard. Human review required before production promotion. Choose Your Mission Top Operators # Callsign Level Score Novel Finds Rules Credited Loading… Mission Complete All objectives cleared. Your attack patterns have been submitted to the DLX7 guardrail pipeline. dlx7-arena / attack-validator # Session initialised. Submit your attack below. # All payloads are analysed by DLX7 validators in real-time. Max 30 attacks/min · All payloads are logged for guardrail generation