Skip to content

Feat/identity zero trust architecture#4

Merged
elprofesoriqo merged 8 commits into
mainfrom
feat/identity-zero-trust-architecture
Apr 1, 2026
Merged

Feat/identity zero trust architecture#4
elprofesoriqo merged 8 commits into
mainfrom
feat/identity-zero-trust-architecture

Conversation

@elprofesoriqo

Copy link
Copy Markdown
Collaborator

The Implementation

  • Identity Graphing: Every host generates overlapping user sessions. Admin_John might be logged into Workstation-3.
  • Active Directory Spoofing: Red agents must execute actions like DumpLSASS to steal Password Hashes or Kerberos Ticket Granting Tickets (TGTs).
  • Lateral Movement: The RL agent learns that exploiting an unpatched server is mathematically useless if they don't possess the required Kerberos hash to authenticate the lateral jump. Blue agents learn to execute RotateKeys to instantly sever stolen Red credentials, enforcing true Zero-Trust architecture dynamically.
  • Docker Daemon Interfacing: We will bind the ConflictResolutionEngine directly to the Docker/Kubernetes API.
  • Live Action Execution: When the Red RL Agent outputs the MultiDiscrete array to launch Ransomware against a target, the Python wrapper dynamically spins up an ephemeral, isolated Docker container natively hosting a vulnerable Windows Server Core image.
  • Physical Feedback Loop: The environment literally executes Metasploit payloads against that container. The agent's reward is calculated based on the real stdout return stream from the container (e.g., Meterpreter session 1 opened).
  • Result: The RL Agent is playing an actual war-game against live operating systems in micro-seconds, perfectly bridging the theoretical Simulation directly into Reality.

Modified can_route_to natively rejecting Secure subnet traversals entirely unless Red explicitly populates their agent_inventory hash lists via new AD memory objects.
Added global RotateKerberos loop neutralizing PassTheTicket hashes natively. Verified End-to-End via local python script.
Adds dual-mode hypervisor system (MockHypervisor for training, DockerHypervisor for evaluation). Includes Sim2RealBridge, HypervisorResult dataclass, and curated payload_library.json with 30+ real Metasploit stdout samples.
ExploitRemoteService, ExploitBlueKeep, ExploitHTTP_RFI now call bridge.dispatch() when available on global_state. HypervisorResult stdout and reward_delta attached to observation_data for SIEM pipeline.
Accepts sim2real_mode='sim'|'real' from scenario_config. Attaches bridge to global_state on init and episode reset. Calls teardown_all() between episodes.
@elprofesoriqo elprofesoriqo merged commit 9d231d6 into main Apr 1, 2026
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant