Your certificate inventory is wrong. Find out by Friday.
Ask three people for the list of certificates in your estate and you will get three lists. None of them is complete, and the gaps between them are where the 3 a.m. outages live. Here is how to find the gaps in a week.
Why every list is wrong
A certificate inventory is not one thing. It is the overlap of three views that see different parts of the world, and every team keeps only one of them.
- The transparency logs record every public certificate that Chrome accepts, because it refuses certificates that are not logged.1 Anyone can search them by name.2 They show what was issued for your domains, including the certificate a contractor requested from a personal account. They show nothing about private certificates, and nothing about where anything is deployed.
- A network scan connects to your address ranges and records what each port presents: public and private certificates alike, including the one on the appliance nobody remembers. It only sees what a host presents on a port, so a certificate sitting in a file, a PKCS#11 token or the Windows certificate store never appears.3
- Host agents see the rest: keystores, files, client certificates, the private key that should not be there. They cost an install per host, so they cover a sample, not the estate.
The week
- Monday: the logs. Pull every certificate in the transparency logs for your registered domains, current and expired. Expect surprises: forgotten subdomains, certificates from authorities you do not use, wildcards you did not know existed.
- Tuesday: the scan. Scan your address ranges on the TLS ports you know and the ones you do not (8443, 9443, database ports, management interfaces). Record the certificate, the name it presents, the expiry, and the fingerprint. Fingerprints matter: the same certificate served from twelve places is one certificate with twelve deployments.
- Wednesday: the hosts. Put an agent on a representative sample: one of each platform, the build servers, the jump hosts. What you find there tells you how wrong the other two lists are.
- Thursday: reconcile. Join the three lists on fingerprint, and match log entries on issuer and serial number, because a log may hold only the precertificate, whose fingerprint differs. Everything in the logs but not on the wire is either dead or deployed somewhere you did not scan. Everything on the wire but not in the logs is private, and needs a private authority behind it that you also need to find. Everything on a host but not on the wire is the long tail.
- Friday: owners. Every certificate gets a person or a team, an expiry, and a triage state. "Owner: unknown" is not an inventory row, it is a finding.4
What you will have
Not a perfect inventory. A first honest one, with a number next to each of the three views so you know how far apart they are. That number is your discovery debt, and it is what the 47-day schedule turns into outages: a certificate you do not know about cannot be renewed by anyone.
trstctl runs the three views as ongoing discovery sources rather than a one-week project: network discovery from an enrolled relay, agent-based discovery from inside hosts, and transparency-log monitoring, all landing in one inventory where an identity with no accountable owner is flagged until someone owns it, and findings carry an event-sourced triage state.3,4
The discovery and inventory page describes the network and host sources, ownership and triage; the observability page covers transparency-log monitoring; and the console demo shows the result on sample data.
Sources
- Google Chrome, Certificate Transparency Policy: publicly trusted certificates must be CT compliant to validate in Chrome.
- IETF, RFC 9162: Certificate Transparency Version 2.0; a public search interface is crt.sh.
- trstctl documentation, Discovery and inventory: network discovery from an enrolled relay, agent-based discovery, and what a scan cannot see.
- trstctl documentation, Discovery and inventory: ownership attribution and discovery triage.