{"data":{"kind":"file","path":"README.md","version_id":"oiu4ijp1d5j897sar01r7psg","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":1644,"modified_at":"2026-05-11T13:37:54.544000","content_hash":"7db854e6eef328e883ef1fc8fdf70bc98b369640c293be9a3292a5c8d232b1ed"},"entries":[],"content":"# Cybersecurity Threat Identification\n\nA Prime Intellect Verifiers environment for evaluating LLM capabilities on cybersecurity threat identification, attack classification, and mitigation suggestion.\n\n## Overview\n\nThis environment tests the model's ability to analyze real-world security scenarios including:\n\n- **SQL Injection** — database attack via malicious queries\n- **XSS** — cross-site scripting attacks in web applications\n- **Phishing/BEC** — social engineering via email\n- **DDoS** — distributed denial of service attacks\n- **MITM** — man-in-the-middle and ARP spoofing\n- **Privilege Escalation** — unauthorized access elevation\n- **Ransomware** — file encryption and extortion\n- **Social Engineering** — USB attacks, impersonation\n- **Supply Chain** — vulnerable dependencies\n- **DNS Tunneling** — data exfiltration via DNS\n- **Container Security** — misconfigurations\n\n## Scoring\n\nThe rubric combines four reward functions:\n\n1. **Keyword Matching** (25%) — required security terms in response\n2. **Attack Classification** (30%) — correct attack category identification\n3. **Severity Assessment** (20%) — correct severity level (LOW/MEDIUM/HIGH/CRITICAL)\n4. **Mitigation Check** (25%) — presence of actionable countermeasures\n\n## System Prompt\n\nThe system prompt instructs the model to:\n- Identify the attack type\n- Rate severity\n- Point to specific evidence\n- Suggest at least 2 countermeasures\n\n## Usage\n\n```bash\nprime env install cybersecurity-threat-id\nprime eval run cybersecurity-threat-id -m <model>\n```\n\n## Tags\n\n`cybersecurity`, `threat-detection`, `single-turn`, `security`, `classification`\n","encoding":"utf-8","truncated":false,"total_bytes":1644},"status":null}