When an incident hits, the first hour decides whether you have evidence or a memory. Here's the triage workflow I actually use — the same one I packaged into a downloadable kit.
The volatile-order rule
RFC 3227 exists because evidence decays fastest in a specific order:
- CPU registers / memory — gone in seconds on shutdown
- Network connections — minutes
- Running processes — minutes to hours
- Disk — the only thing that survives
Skip the order and you've already contaminated the scene.
The 30-minute triage
Minute 0–5: Snapshot, don't touch. Capture ps auxww, netstat -tulpen (or ss -tulpen), mount, last -20, and running cron/systemd timers. Read-only commands only — every command you run writes to the disk you're trying to preserve.
Minute 5–15: Network first. Established connections tell you what the box is doing right now, not what it did. ss -tnp plus lsof -i maps every socket to a process. Unexpected outbound 443s to rare ASNs are your fastest lead.
Minute 15–25: Persistence check. The three places everything hides:
-
systemctl list-unit-files --state=enabled— new services -
crontab -lfor every user +/etc/cron.*— new jobs -
~/.ssh/authorized_keys,~/.bashrc,/etc/ld.so.preload— the classics
Minute 25–30: Hash and document. sha256sum every captured file before analysis. Chain of custody starts at capture, not at report time — if you can't prove the evidence didn't change, a defense attorney will say it did.
What I packaged
The full kit — RFC-3227-ordered checklist, a tested triage_capture.sh that grabs 18 evidence artifacts and writes a sha256 manifest automatically, and a chain-of-custody form:
Store (pay in BTC, no account, instant download): https://e72fd9b2dcb05c.app.lhr.life
The script is the part worth the $14 alone — it encodes the ordering so you can't forget step 3 while panicking at 3am.
Running incident response on a team? What does your triage checklist look like — and does it survive contact with a real 3am page?
Top comments (0)