PKI & Certificates
How asymmetric keys and a chain of trust let strangers prove identity.
By the end of this lesson, explain certificate chains, TLS identity, and mutual TLS for service identity.
Public Key Infrastructure (PKI) is how two parties who have never met can trust each other's identity. A certificate binds a public key to a name and is signed by a Certificate Authority (CA). If you trust the CA, you trust every certificate it signs — a transitive chain of trust.
First, see the whole system. Then we’ll open it up.
Chain of trust
- A root CA signs an intermediate CA, which signs a leaf certificate for a service.
- A client that trusts the root therefore trusts the leaf via the chain.
flowchart TB root[Root CA] -->|signs| inter[Intermediate CA] inter -->|signs| leaf[Leaf cert: service.example.com] client[Client] -->|trusts root| root client -->|therefore trusts| leaf
In TLS, the server presents its certificate chain; the client verifies the signatures up to a trusted root and checks the name matches. Mutual TLS (mTLS) adds the reverse: the client also presents a certificate, so both sides authenticate. mTLS is the backbone of service-to-service identity in meshes.
This is where the abstraction starts leaking.
The hard part of PKI is not signing — it's rotation, revocation, and clock. Certificates expire (an outage if you forget to renew), revocation (CRL/OCSP) is famously unreliable, and a wrong clock rejects valid certs. Short-lived certs sidestep revocation by expiring before it matters.
Expired or misissued certificate
- Trigger
- A certificate lapses, or a CA is tricked into issuing one for a name it shouldn't.
- Symptom
- Sudden connection failures (expiry) or silent impersonation (misissuance).
- Blast radius
- Outage across everything that trusts the cert, or MITM of the affected name.
- Mitigation
- Automate renewal; short lifetimes; CAA records; certificate transparency monitoring.
Deep dive: Clock Skew
Long-lived certs are easy to issue but hard to revoke and dangerous when leaked. Short-lived certs (hours) need automation but make revocation almost free. Modern workload identity chooses short-lived + automated.
Why do short-lived certificates reduce the need for revocation?
A certificate that expires in hours limits the window a leaked one is usable, so it self-revokes before CRL/OCSP would matter.