Security Glossary

Using asymmetric cryptography, the problem of key distribution becomes one of authentication; i.e. how to make sure that a given key really does belong to the entity that claims to own it.

• providing trusted ‘root’ certificates to users (End Entities), by supplying them with the CA’s public key via out-of-band means.
• certifying End Entities by generating and distributing certificates for them. The certified EE is the subject of the certificate: the CA is the issuer.
• supporting certificate revocation and revocation checking: if an EE suspects that their key has been compromised, they contact the CA which issued it, who should revoke their certificate.

A CA will always have a root certificate-signing key pair, which must be authenticated to End Entities via out of band channels. This key pair is not logically certified by anything, but it is usually distributed inside a self-signed certificate to afford some degree of tamper evidency.

However, CAs do not have to use their root key pair to issue certificates directly to End Entities. For organizational reasons and to reduce the exposure of keys, a CA may have a single root signing key pair, which it uses to certify a set of subordinate key pairs, which in turn are used to certify End Entities. Also, CAs may certify the signing keys of other CAs by issuing cross certificates, which enable interoperation between two distinct PKIs.

Digital signatures have two primary uses: to demonstrate someone's identity by signing some challenge, as in client authentication in TLS, in which the client signs a hash of the messages that have been exchanged, and more strongly, for someone to demonstrate their acceptance of some human-processable information (e.g. 'Please withdraw £10 000 from my bank account') as in the WMLScript Crypto API SignText function.

A common example of an out of band channel is a motorcycle courier.

A way of modelling real-world trust relationships which enables users of public key cryptography to have confidence in the ownership of the public keys they are using. A PKI consists of:

• a trusted third party
• an out of band means of distributing the TTP's public key certificate to relying parties
• a means of distributing other certificates to relying parties
• arrangements for the revocation and renewal of these certificates
• certificate management and validation software on the relying party's computer

The TTP uses its signing key pair to create certificates for other entities, which relying parties can use to authenticate these other entities.

We can classify PKIs according to whether they are hierachical or flat. In hierachical PKIs, such as the one defined in the PKIX set of standards, there is a distinction between users of the PKI such as End Entities and Relying Parties, and entities responsible for issuing and distributing certificates such as CAs and RAs. In a flat PKI such as the web of trust underpinning PGP, there are no entities whose sole role is to issue certificates; instead users of the PKI certify each other.

Symmetric cryptography is fast, but suffers from the problem of how to distribute the key privately. Asymmetric cryptography is an attempt to alleviate the key distribution problem, by reducing the requirement for the distributed key from one of privacy to one of authentication.

In this protocol a client connects to a server; the two then perform a handshake in which they exchange a symmetric key by using asymmetric cryptography, which is then used to encrypt their communications, providing the secrecy element.

Without the authentication element secrecy is not very useful; although only client and server can understand the data exchanged, the client doesn't know who the server is or vice versa. TLS provides the capability for server authentication, in which the client establishes who the server is, and client authentication in which the server establishes who the client is.

A CA is a type of TTP.