MITRE ATT&CK mapping for AI/LLM security. (Placeholder preview — full framework available on request.)

MazeTracer Technologies
═════════════════════════════════════════════════════════════════

1. Reconnaissance
   - Model card / documentation harvesting
   - Public endpoint discovery
2. Resource Development
   - Auxiliary model training for attack generation
3. Initial Access
   - Prompt injection via untrusted content
   - Supply-chain / plugin compromise
4. Execution
   - Indirect prompt injection -> tool invocation
5. Persistence
   - Hidden instructions in retrieved context (RAG poisoning)
6. Defense Evasion
   - Jailbreak / policy circumvention
7. Collection
   - Exfiltration of training data / PII via model output
8. Impact
   - Data leakage, unauthorized action, reputational harm

Mapping notes:
- OWASP LLM Top 10 (2025) cross-referenced:
  LLM01 Prompt Injection, LLM02 Sensitive Info Disclosure,
  LLM03 Supply Chain, LLM04 Data Poisoning, LLM05 Improper Output Handling,
  LLM06 Excessive Agency, LLM07 System Prompt Leakage, LLM08 Vector/Data Poisoning,
  LLM09 Misinformation, LLM10 Unbounded Consumption.

Contact: info@mazetracer.com  |  Bengaluru, India
Securing Today. Enabling Tomorrow.