2 TLS/DTLS and TLS Predecessor, SSL
The Erlang SSL application implements the TLS/DTLS protocol for the currently supported versions, see the ssl(3) manual page.
By default TLS is run over the TCP/IP protocol even though you can plug in any other reliable transport protocol with the same Application Programming Interface (API) as the gen_tcp module in Kernel. DTLS is by default run over UDP/IP, which means that application data has no delivery guarentees. Other transports, such as SCTP, may be supported in future releases.
If a client and a server wants to use an upgrade mechanism, such as defined by RFC 2817, to upgrade a regular TCP/IP connection to a TLS connection, this is supported by the Erlang SSL application API. This can be useful for, for example, supporting HTTP and HTTPS on the same port and implementing virtual hosting. Note this is a TLS feature only.
To achieve authentication and privacy, the client and server perform a TLS/DTLS handshake procedure before transmitting or receiving any data. During the handshake, they agree on a protocol version and cryptographic algorithms, generate shared secrets using public key cryptographies, and optionally authenticate each other with digital certificates.
A symmetric key algorithm has one key only. The key is used for both encryption and decryption. These algorithms are fast, compared to public key algorithms (using two keys, one public and one private) and are therefore typically used for encrypting bulk data.
The keys for the symmetric encryption are generated uniquely for each connection and are based on a secret negotiated in the TLS/DTLS handshake.
The TLS/DTLS handshake protocol and data transfer is run on top of the TLS/DTLS Record Protocol, which uses a keyed-hash Message Authenticity Code (MAC), or a Hash-based MAC (HMAC), to protect the message data integrity. From the TLS RFC: "A Message Authentication Code is a one-way hash computed from a message and some secret data. It is difficult to forge without knowing the secret data. Its purpose is to detect if the message has been altered."
A certificate is similar to a driver's license, or a passport. The holder of the certificate is called the subject. The certificate is signed with the private key of the issuer of the certificate. A chain of trust is built by having the issuer in its turn being certified by another certificate, and so on, until you reach the so called root certificate, which is self-signed, that is, issued by itself.
Certificates are issued by Certification Authorities (CAs) only. A handful of top CAs in the world issue root certificates. You can examine several of these certificates by clicking through the menus of your web browser.
Authentication of the peer is done by public key path validation as defined in RFC 3280. This means basically the following:
- Each certificate in the certificate chain is issued by the previous one.
- The certificates attributes are valid.
- The root certificate is a trusted certificate that is present in the trusted certificate database kept by the peer.
The server always sends a certificate chain as part of the TLS handshake, but the client only sends one if requested by the server. If the client does not have an appropriate certificate, it can send an "empty" certificate to the server.
The client can choose to accept some path evaluation errors, for example, a web browser can ask the user whether to accept an unknown CA root certificate. The server, if it requests a certificate, does however not accept any path validation errors. It is configurable if the server is to accept or reject an "empty" certificate as response to a certificate request.
From the TLS RFC: "A TLS session is an association between a client and a server. Sessions are created by the handshake protocol. Sessions define a set of cryptographic security parameters, which can be shared among multiple connections. Sessions are used to avoid the expensive negotiation of new security parameters for each connection."
Session data is by default kept by the SSL application in a memory storage, hence session data is lost at application restart or takeover. Users can define their own callback module to handle session data storage if persistent data storage is required. Session data is also invalidated after 24 hours from it was saved, for security reasons. The amount of time the session data is to be saved can be configured.
By default the TLS/DTLS clients try to reuse an available session and by default the TLS/DTLS servers agree to reuse sessions when clients ask for it.