Files
Steffen Skui 2ff4e703c4
Build and push soc-collector / build (push) Successful in 5s
cve: exclude the collector's own container from the scan (+ SOC_EXCLUDE_IMAGES)
2026-06-25 10:31:40 +02:00

3.5 KiB

soc-collector

A single, run-anywhere container that scans your estate and pushes the results to a SOC Center instance as auto-updating reports. No JSON to write, no curl — you give it an API endpoint and a token, point it at some targets, and run it. Same report slug every run, so SOC Center versions each scan, diffs what changed, and alerts you only on new findings.

Run it

docker run --rm \
  -e SOC_URL=https://soc.ethica.no \
  -e SOC_TOKEN=soc_xxxxxxxx \
  -e SOC_SLUG=my-server \
  -e SOC_TARGETS=10.0.0.5,10.0.0.6 \
  -e SOC_DOMAINS=soc.ethica.no,ethica.no \
  -v /var/run/docker.sock:/var/run/docker.sock:ro \
  git.skui.io/ethica/soc-collector:latest

Each run produces one report (named by SOC_SLUG) merging all the sources below. One server = one report; run it per server with a distinct SOC_SLUG for one report each.

It runs once and exits. To keep reports fresh, schedule that command — e.g. nightly:

0 2 * * *  docker run --rm -e SOC_URL=… -e SOC_TOKEN=… -e SOC_TARGETS=… -e SOC_DOMAINS=… \
             -v /var/run/docker.sock:/var/run/docker.sock:ro git.skui.io/ethica/soc-collector:latest

The SOC_TOKEN is a write-scoped token you mint once in SOC Center (Admin → Tokens). Reports are owned by, and visible per, that token's user.

What it collects

One run = one consolidated report (slug = SOC_SLUG) merging whichever of these sources have their inputs, criticals first:

Source What it checks Needs
surface nmap open-port / service scan, flags risky exposed ports SOC_TARGETS
cve trivy of every running image — fixable HIGH/CRITICAL only, grouped by package Docker socket
tls cert-expiry, weak TLS, missing security headers SOC_DOMAINS
config 0.0.0.0 port binds, --privileged, rw docker-socket mounts Docker socket

Each source self-skips when its inputs are absent — no SOC_TARGETS → no nmap; no socket mount → no Trivy / config. So the same image is safe to run anywhere; you just enable what you give it.

Configuration (all via env)

Var Required Default Meaning
SOC_URL SOC Center base URL, e.g. https://soc.ethica.no
SOC_TOKEN write-scoped API token
SOC_TARGETS for nmap comma/space hosts or IPs to scan
SOC_DOMAINS for tls comma hostnames to TLS-probe
SOC_SOURCES surface,cve,tls,config which sources to run
SOC_SLUG soc-scan the report's slug — set one per server (e.g. hetzner-prod)
SOC_TITLE Security scan — <slug> the report's title
SOC_VISIBILITY private private or authenticated
SOC_TAGS collector extra report tags
SOC_EXCLUDE_IMAGES comma-separated substrings; running images matching any are skipped by the cve source

Notes

  • Docker socket is mounted read-only (:ro) and is only used to list running containers and inspect their config. For extra isolation, point it at a scoped socket-proxy instead of the raw socket.
  • The cve source skips the collector's own container, so the scanner never reports on itself. Use SOC_EXCLUDE_IMAGES to drop any other images you don't want scanned.
  • The collector is pure-stdlib Python; the image just adds nmap and trivy.
  • Multi-server: run one container per server with a distinct SOC_SLUG — one report each.

License

MIT — see LICENSE.