DungeonQ
Divert suspicious sessions into persistent decoy worlds
Start new thread

A decoy outage must not reopen the protected origin

by
•

DungeonQ's defensive deception runtime keeps designated sessions working inside persistent synthetic worlds. What happens when that path fails matters as much as the successful walkthrough.

In the self-hosted reference, a synthetic transport failure does not fall back to the protected origin. Invalid authority, missing evidence, or unreadable state return bounded errors. The gateway records an attempt before dispatch, so an interrupted operation cannot simply disappear from the evidence.

The published September 18 acceptance record includes a useful negative check: with the containers stopped, verification returned INCONCLUSIVE instead of reusing an earlier passing result.

For operators, preserve the failed run's records and inspect the uncertainty. Automatic reconciliation of unknown outcomes is not implemented; deleting state is not a recovery procedure.

These are documented mechanisms and historical checks of an owned reference using artificial resources, not a new release or proof of production protection or AI deception.

Operation guide:

Recorded acceptance and limits:

33 views

Add a comment

Replies

Best

Not falling back to the real origin when they decoy breaks is the detail that stood out to me.

Thanks, Victor. That failure path is central to the design. DungeonQ is intended to buy security teams time and room to respond by keeping a diverted session inside a persistent decoy world. An outage must preserve that boundary so a human responder or an authorized AI agent can investigate and decide what happens next.

INCONCLUSIVE is much mor honest result than pretending an older passing run still counts.

Thanks, Andrea. An operator needs to know what the evidence supports right now. INCONCLUSIVE makes the uncertainty visible, so a human responder or an authorized AI agent can investigate before choosing the next action. Creating that space for informed response is a core goal of DungeonQ.

I like that the attempt gets recorded before dispatch. That could save some head scratching after an interrupted run.

Thanks, Filipe. Recording the attempt before dispatch gives the responder a starting point when execution is interrupted. DungeonQ is designed to create time for investigation; that time is more useful when people or authorized AI agents can reconstruct what was attempted and what remains unknown. The record alone does not resolve an interrupted operation.

The bounded errors for missing or unreadable state makes sense. Better than quietly trying something else.

Thanks, Zarnappa. Missing state should leave the next decision with the defender. Bounded errors make the problem visible while keeping the session within its assigned boundary. The aim is to give a security operator or an authorized AI agent room to assess the situation and choose a controlled recovery.

The note about this not being proof of production protection is useful. There's a pretty big gap between a refrence check and real world deployment.

Agreed, James. The gap needs to stay explicit. DungeonQ's goal is to buy security teams time and room to observe suspicious activity and respond, with a human or authorized AI agent able to take over. We make no promise of stopping every attack. The current reference checks establish specific behavior under tested conditions; real deployment needs its own integration, isolation, and operational validation.

The “INCONCLUSIVE” result is actually a useful approach. It’s better to clearly show uncertainty than reuse an older successful result and give a false sense that everything is fine.