Security / ARTICLE

Coldframe: A Honeypot That Poses as a Supermarket Fridge Controller

Coldframe is a single-file Python honeypot that impersonates Danfoss and Dixell refrigeration controllers, logging credential attempts and exploit traffic.

Matching Real Controller Fingerprints

Coldframe is a Python standard library honeypot that impersonates two supermarket refrigeration controllers: the Danfoss AK-SM 800A system manager and the Copeland/Dixell XWEB EVO supervisor. The project builds on eppser/danfoss-ak-sm-decoy and public research from Synacktiv, Claroty, and Ironhackers. The README describes the context: in 2026, attackers used default passwords and unpatched firmware to break into internet-facing refrigeration supervisors across Europe and force defrost cycles. Coldframe logs who connects, which credentials are tried, and which public exploits appear, without offering any real control plane.

The Danfoss persona reproduces scanner-visible attributes of real AK-SM 800A units: the nginx server header and header order, the “Danfoss StoreView” page title, TLS 1.2 only, and a one-year self-signed certificate with CN=DanfossSM800A. Each build generates fresh keys. Weak logins such as Supervisor / 12345 succeed, as does a correct CVE-2025-41450 “code of the day.” A successful login returns a session token and a dashboard of fake refrigerated cases with defrost buttons, and every action is logged as a control event.

The Dixell persona follows the same approach for XWEB EVO: no Server header, the same security-header set and order, matching asset paths and page size. Login uses the real /cgi-bin/login.lua flow; the factory Admin / Admin credentials work, five failures in five minutes lock the user, and attempts to use the XWEB Pro auth_mode bypass (CVE-2026-25085) are tagged.

The repository does not include the vendor favicon or the Danfoss start page’s Angular loader because they are proprietary; private deployments can point FAVICON_FILE and INDEX_FILE at copies they are licensed to use, allowing the decoy to serve the loader byte for byte.

Exploit Detection and Logging

Coldframe adds an exploit field to matching requests. It recognizes Danfoss CVE-2023-25914 path traversal via get_file, CVE-2023-25915 command injection via load_cert, CVE-2025-41452 nginx header overwrite via load_nginx_headers, and the CVE-2025-41450 “code of the day” bypass on getauth with eds_only or quantum_only parameters. For that bypass, the log notes whether the submitted code is correct. On the XWEB side, it flags CVE-2021-45420 upload and command paths under /cgi-bin/ and the CVE-2026-25085 auth_mode bypass.

Events are written as JSON lines to events.jsonl. Request entries include method, path, Host, User-Agent, Authorization, and body. Other event types cover login attempts, XML API commands, control actions, TLS errors, and Modbus activity. A summary.py script converts logs into short reports.

Both personas can optionally run a Modbus TCP service with fake case temperatures, setpoints, and a defrost register. Real AK-SM 800A units use Modbus only to talk to their own case controllers, so Modbus is off by default; enabling it can catch generic industrial scanners.

Deployment and Limits

Quick start is a single command: PERSONA=danfoss HTTP_PORT=8080 python3 -I honeypot.py. For production, a cloud-init config sets up Ubuntu 24.04 and Podman. The container runs with rootful Podman, –userns=auto, a read-only filesystem, and no capabilities. Log fields are treated as untrusted input, and connections, request bodies, and log lines are capped. Memory is limited to 128 MB and process IDs to 512. Connections that trickle data are cut after 30 seconds, and refused connections are logged. Over-long log lines are truncated, and summary.py neutralises terminal escape codes from attacker input.

The README lists limitations. Interaction after login is shallow, with only a small dashboard and a handful of XML commands. Public code can be fingerprinted, and some static paths and XML response shapes are plausible rather than verified. Cloud IP addresses may cause scanners to flag the device as a likely honeypot. Two early instances saw only generic scanners and Shodan’s refrigerator.census.shodan.io crawler.

END
Feedback