X.509 client certificate authentication lets a device prove who it is with a private key that never leaves it, instead of a shared password. In the TLS handshake the server asks for a certificate, the device sends its certificate chain and signs the handshake with its key, and the server checks that the chain leads to a certificate authority it trusts.
Most comparisons of certificates and passwords assume a person at a browser. In a fleet of field devices, when the same password is flashed onto every unit, one unit taken from a site gives away the secret of all of them, and changing it means touching every device. I covered the same failure for controllers in water utility cybersecurity.
Client certificate vs password authentication: what changes for devices?
With a certificate, the device never sends its secret anywhere; it proves it holds the private key by signing. TLS 1.3 (RFC 8446) calls this signature CertificateVerify and requires a client to send it whenever it authenticates with a certificate.
| Question | One shared password | One certificate per device |
|---|---|---|
| What crosses the network | The password itself, inside TLS | A signature; the private key stays on the device |
| One unit is stolen | Every unit is exposed | Revoke that one certificate |
| Rotation | Touch every device at once | Each device renews on its own schedule |
| Who connected | The logs cannot tell units apart | The certificate names the device |
Per-device passwords fix the stolen-unit row, but the secret still travels at every login and still has to be stored and rotated. I would rather spend that effort on a small PKI.
What is the difference between a client and a server certificate?
Both are X.509 certificates in the same format; they differ in purpose and in what the other side checks. RFC 5280 defines extended key usage values for TLS server authentication (id-kp-serverAuth) and TLS client authentication (id-kp-clientAuth), and a certificate that carries the extension must only be used for the purposes it lists.
In TLS 1.3 the server always sends its certificate when the key exchange uses certificates. The client sends one only when the server asks with a CertificateRequest. The server certificate names the host the device connects to; the client certificate should name the device, for example by its serial number, so the server can tie the connection to one unit.
Is PKI practical for IoT devices?
Yes, once issuing and renewing certificates needs no person at the device:
- The device generates a key pair; the private key is never exported.
- It sends a certification request (PKCS #10, RFC 2986) carrying its name and public key, signed with its private key.
- The certificate authority checks who is asking and returns a certificate with a fixed lifetime.
- The device renews before expiry, and a lost or retired unit is revoked.
RFC 5280 describes revocation through certificate revocation lists: signed, time-stamped lists of revoked serial numbers that a relying system checks along with the signature and validity dates. A device also needs a working clock, or it cannot check the validity dates of the server’s certificate.
Hermetry connects field devices to live dashboards, rules and alerts: IoT telemetry and SCADA in one platform. Its licensed devices log in with X.509 client certificates from Hermetry’s own device certificate authority: the device makes its own key pair, sends a signing request and receives a certificate whose common name is the device serial. At every login Hermetry checks the chain, revocation, expiry and that the common name matches the serial and client ID. Lifetimes are 365 days by default, and the device certificates page covers renewal and revocation. Third-party devices use access tokens or MQTT username and password instead.
Certificates prove identity. Which systems a device may reach is a network question, covered in IoT network segmentation for plants, and the wider set of device requirements in IEC 62443 explained.
Frequently asked questions
Does a client certificate replace checking the server certificate?
No. Mutual TLS means both sides authenticate. The device must still verify the server’s certificate chain and name, or it could send its data to an impostor that simply accepts any client. The client certificate adds the second direction: it lets the server know which device is on the line.
What happens when a device certificate expires?
Certificate path validation in RFC 5280 checks the validity period, so a server that validates correctly refuses an expired client certificate. Devices should renew well before expiry, and the fleet should report certificates that are close to expiring, so a forgotten unit is caught before it drops off.
Can several devices share one certificate?
Technically yes, but it brings back every problem of a shared password. The private key sits on every unit, revoking it cuts off the whole group, and the server cannot tell the devices apart. Give each device its own key pair and certificate.

