Root CA (self-signed, the trust anchor) → intermediate or issuing CA → end-entity certificate. At least two signatures are checked: the one on the intermediate and the one on the server certificate. Trust rests on the root being installed as a trust anchor.

The three roles:

What gets verified: the signature on certificate 3 is checked with the public key in certificate 2, and the signature on certificate 2 with the public key in certificate 1. That makes two signatures. The root's self-signature proves nothing by itself, since anyone can sign their own certificate.

Why the root is trusted: it is stored in the operating system or browser as a trusted root CA, which makes it the trust anchor. Validation always ends there.

Installing a new root by hand: before accepting an unknown root CA certificate, compare its fingerprint (a hash of the certificate) with the fingerprint from an independent, trustworthy source, such as the CA's website or a phone call. Otherwise an attacker could slip in their own root and every certificate they issue would look valid.

An insider is a legitimate participant (Alice or Bob) who behaves dishonestly; an outsider (Eve) is a third party attacking the communication channel.

Outsider (Eve) can perform:

• Eavesdropping, Modifying, Inserting, Replaying, Deleting, Masquerade

Insider attacks — only a participant can do these:

• Alice can deny having sent a message (non-repudiation of origin)
• Bob can deny having received a message (non-repudiation of receipt)

This distinction matters because insider attacks require fundamentally different countermeasures. You can't protect against a dishonest sender with a symmetric shared secret — both parties know the key. This is exactly why digital signatures (asymmetric) exist: they provide non-repudiation that symmetric MACs cannot.

Go deeper:

• Non-repudiation — why a dishonest insider needs signatures, not a shared secret.