V1 · Authorization
Is the action permitted by policy?
VERIFIED AUTONOMOUS DEFENSE · v0.9.0-expo-rc2
INVERIQ
AI decides. INVERIQ verifies.
INVERIQ validates AI-generated cybersecurity actions before safely enforcing them across IoT, Edge, Cloud, and OT infrastructures.
Core control loop
INVERIQ is built around a simple principle: no AI-generated security action should reach infrastructure blindly.
Detect→Decide→Verify→Defend→Verify
V2 · Target
Does the action match the observed evidence and intended target?
V3 · Reachability
Will critical flows remain reachable?
V4 · Availability
Will protected services remain available?
V5 · Blast Radius
Is collateral impact within the configured threshold?
V6 · Security Effectiveness
Is the action expected to reduce or stop the attack?
STATIC VERIFICATION DEMO
Evaluate a candidate mitigation
Describe a candidate action. This front-end illustrates INVERIQ's all-gates verification principle; it does not run the actual engine or enforce firewall rules.
Safety architecture
AI recommendation→Structured action→Deterministic verification→Trusted enforcement adapter
The optional LLM does not write directly to infrastructure. Candidate actions must pass deterministic verification before enforcement.
Current demo focus
The repository's primary IoT Tech Expo scenario is credential brute force mapped to MITRE ATT&CK T1110. A deliberately unsafe mitigation is rejected, while a verified safe candidate can be enforced with nftables inside the isolated Docker lab. Real enforcement is confined to that lab; the host firewall is not modified by the demo tooling.