SECURITY_THREAT_MODEL.md

AIDomesticCoreAIJ – Security Threat Model

Threat Actors

• Malicious users
• Compromised AI models
• Rogue tools or plugins
• Insider threats

Attack Surfaces

• Model inference interfaces
• Tool execution runtime
• Memory storage backends
• Configuration and policy injection

Key Threats

• Prompt injection
• Data exfiltration
• Privilege escalation
• Model poisoning
• Unauthorized tool execution

Mitigations

• Strict sandboxing
• Least-privilege access
• Deterministic execution modes
• Audit logging and replay
• Policy enforcement at kernel level