Key features¶
Decoys¶
- Five decoy roles, each a believable server with only that server's ports: Windows file server, Windows domain controller, Linux web hosting, FTP/backup server and Linux DNS server. See Decoy roles.
- You choose the roles (up to 3 decoys in Community Edition), their host names and their network interfaces. Each decoy answers on its own interface's IP and MAC address, like a real server.
- Your identity: the decoys carry your AD/DNS domain. The company name, NetBIOS name and mail domain are derived from it, and users are picked from a built-in name pool per installation.
- Per-install randomness: passwords, keys, users, dates and file contents differ on every installation, so there is no shared fingerprint between ReD-Pot deployments.
- Realistic protocol behaviour: SMB with NTLM logons, RDP with NLA, LDAP/LDAPS with NTLM binds, a Kerberos KDC that issues tickets for the bait accounts, Microsoft-style DNS, an Apache-like web server, vsftpd-like FTP, and an SSH server with a simulated shell. Nothing an attacker sends is executed. See Protocol behaviour.
Baits and honeytokens¶
- Baits generated for your domain:
passwords.txt, SSH deploy keys,.envfiles, database dumps, a Group Policy Preferences password, a Kerberoastable service account, and Word documents that report when they are opened. - The baits form a breadcrumb chain: credentials on one decoy unlock the next, so an intruder who takes the bait leaves a trail. See Breadcrumb chain.
- Every bait is a honeytoken with a unique id. When it is used, ReD-Pot knows what was stolen, where from, who used it and where.
Detection and investigation¶
- Severity ladder L1 touch to L5 honeytoken, with escalation for port scans, password spraying and path brute force.
- Events from one source are grouped into one incident that only goes up the ladder.
- Who did it: workstation name (from SMB, RDP and NTLM), the logged-on user, the user names tried, and the tools seen (user agents, SSH clients).
- Internal network map with animated attack paths, and a world attack map for internet-facing decoys (offline GeoIP).
- Tamper detection: a decoy that stops reporting opens an L4 "possible tampering" incident; a decoy whose services are flooded opens an L4 "possible DoS" incident.
Operations¶
- One interactive installer with back, validation and a review screen, or fully unattended flags.
- Remote sensors in other VLANs or offices over a certificate-pinned TLS link (about 0.6 kbit/s idle).
- SIEM feed:
alerts.json(JSON Lines, rotated and gzipped), optionalevents.jsonwith every event. - Export of incidents and alert records for a date range, from the console or the command line.
- Wall mode for SOC screens with a read-only kiosk link.
- In-place update with backup and automatic rollback (
sudo ./install.sh update). - Lightweight: Python and SQLite, no Elastic stack. About 140 MB RAM for the manager with two decoys.
- Offline after installation: local GeoIP, no cloud services, no telemetry and no outbound alerts.
Hardened by default¶
Decoys run as an unprivileged user in read-only containers, with no capabilities except binding low ports, and with
no-new-privileges. On a decoy's interface only the decoy's own ports answer; the admin ports answer only on the
management interface, to allow-listed networks. See Security hardening.