New: the AI firewall watches the whole platform and explains each incident in plain words. See how it works

Self-signed certificate vs private CA for devices on an internal network

A self-signed certificate is its own trust anchor, so a fleet of them means one anchor per device and no revocation. Why a private CA scales, and what running one takes.

A self-signed certificate vouches only for itself, so every client that should trust it needs that exact certificate installed. A private certificate authority (CA) signs device and server certificates under one root, and clients install only the root. For more than a handful of devices, a private CA wins, if its root key is protected.

On an internal network full of gateways and meters, the browser warning matters less than what each client must trust and what happens when a key leaks.

Why are self-signed certificates not trusted?

Nothing vouches for a self-signed certificate except the certificate itself. RFC 5280 defines it as a certificate whose signature verifies with the public key it carries, and path validation starts from a trust anchor received “by some trustworthy out-of-band procedure”.

You can add it to a client’s anchors by hand. But RFC 5280 says a trust anchor given as a self-signed certificate is not part of the certification path, so RFC 5280 path validation never checks it for revocation. To stop trusting a leaked one, you remove it from every client that holds it.

Self-signed certificate vs private CA: what changes for a fleet?

With self-signed certificates, every device is its own trust anchor. With a private CA, the root is the only anchor.

QuestionSelf-signed per devicePrivate CA
What each client trustsEvery peer’s certificateOne root certificate
Adding a deviceInstall it on every clientThe CA signs it; clients change nothing
One device key leaksRemove it everywhereRevoke it in the next revocation list
Keys to protect (the price of a CA)Each device’s keyEach device’s key plus the CA keys

How do you run HTTPS on an internal network?

Create a private CA, install its root certificate in each client’s trust store, and give each server a certificate that names its host. I would go in this order:

  1. Create the root key and certificate on a machine that is never networked. Set the basic constraints cA flag, or RFC 5280 forbids using the key to verify certificate signatures.
  2. Have the root sign an issuing CA for daily work, and put the root key back offline.
  3. Install the root certificate on every client through a channel you control.
  4. Put each server’s host name in the subjectAltName dNSName entry. RFC 9525 says the Common Name must not be used to identify a service.
  5. Publish a certificate revocation list (CRL) and renew it before the nextUpdate date that RFC 5280 requires on every CRL.

What happens if a CA private key is lost?

A destroyed key stops all new signing; a stolen key lets someone else sign certificates your devices accept.

Certificates from a destroyed issuing CA still verify until they expire, but nothing can sign a new revocation list for them. The root signs a replacement, and devices, which trust the root, change nothing. If an issuing CA key is stolen, the root revokes that CA’s certificate. A stolen root has nothing above it, and every client needs a new root installed. For the same reason I would keep the root key, offline, after signing the issuing CA: without it, the root could neither revoke that CA nor sign its successor.

Hermetry connects field devices to live dashboards, rules and alerts: IoT telemetry and SCADA in one platform. Its licensed devices sign in with certificates from Hermetry’s own device certificate authority, whose root keys stay offline, wrapped under a key split between two custodians. A device issuing CA signs for at most 24 months before a new one takes over; devices trust the root, so they need no change. Third-party devices use device credentials such as access tokens or MQTT username and password.

How a device proves its identity during the handshake is covered in X.509 client certificate authentication, and which systems it may reach in IoT network segmentation for plants.

Frequently asked questions

Can an internal certificate use an IP address instead of a host name?

Yes. Put the address in an iPAddress entry of the subjectAltName extension. RFC 9525 says an IP address reference must match that entry exactly, and RFC 2818 requires the entry when a URL uses an IP address.

What does a device need to check a certificate chain?

The root certificate as its trust anchor, any intermediate certificates, and a correct clock. RFC 5280 path validation checks that each certificate was valid at the time in question, so a device whose clock reset can reject a good certificate.

Can a private CA limit which names it signs for?

Yes, with the name constraints extension. RFC 5280 allows it only in CA certificates, and later certificates in the path must keep their subject names inside the name space it sets. On clients that enforce it, this limits what a stolen issuing key could sign for.

Share

See your field data live this week

Start a trial in your region with your first devices, or book a demo and bring your own use case.