How does a Certificate Signing Request work, and which key is used where?

The applicant signs the request with its own private key to prove it holds the matching key; the CA checks that signature with the public key inside the request, then signs the certificate with the CA's private key.

1. Applicant generates a key pair and builds the request: its public key + the requested subject name and extensions.
2. Applicant signs the request with its private key. This is proof of possession: it shows the request was made by whoever holds the private key matching the enclosed public key, so nobody can ask for a certificate on someone else's public key.
3. CA verifies that signature using the public key in the CSR. It also checks that the requester may have such a certificate (here: AD login to /certsrv and permission on the "Web Server" template).
4. CA issues the certificate: it fills in issuer, serial number, validity and possibly overrides extensions from its template, then signs everything with the CA's private key.
5. Applicant installs the certificate next to its private key (www.cer in /etc/ssl/certs/).
6. Clients later verify the certificate with the CA's public key (from the CA certificate in their trust store).

Keys at a glance: the applicant's private key signs the CSR and stays home; its public key travels in the CSR and ends up in the certificate; the CA's private key signs the certificate; the CA's public key verifies it.

* Which key signs what in a certificate request. *

Go deeper:

• Wikipedia: Certificate signing request — the proof-of-possession signature explained.