Glossary
SRTP (Secure Real-time Transport Protocol)
The encrypted form of RTP. SRTP protects voice and video packets in transit with encryption, integrity checks and replay protection, and every WebRTC call uses it.
General definition
SRTP (Secure Real-time Transport Protocol) is defined in RFC 3711 as a security profile for RTP, the protocol that carries audio and video packets in VoIP and video calls. RTP itself sends media in the clear. SRTP encrypts the payload of each packet, appends an authentication tag so a receiver can detect tampering, and tracks sequence numbers to reject replayed packets. The RTP header stays readable so routers and media servers can still handle the stream. A companion profile, SRTCP, protects the RTCP control packets that report on call quality.
- Confidentiality: the payload is encrypted with AES in counter mode by default; AES-GCM profiles (RFC 7714) combine encryption and authentication in one step
- Integrity: an HMAC-SHA1 tag over header and payload catches modified packets
- Replay protection: a sliding window of seen sequence numbers blocks re-injected packets
- Low overhead: a few bytes per packet and no extra round trips once keys are in place
SRTP does not say how the two sides agree on keys. Classic SIP deployments often use SDES, which carries keys inside the SDP session description and therefore relies on the signaling path being secure. WebRTC instead requires DTLS-SRTP (RFC 5763 and RFC 5764): the endpoints run a DTLS handshake over the media path, verified against certificate fingerprints exchanged in signaling, and derive the SRTP keys from it. Unencrypted RTP is not permitted in WebRTC at all (RFC 8827).
One detail matters for architecture. SRTP protects each hop, not the whole path. When a call passes through an SFU or other media server, that server terminates SRTP on the incoming leg and re-encrypts on each outgoing leg, so it can access the media. True end-to-end encryption of a group call needs an extra layer on top, such as SFrame carried through WebRTC insertable streams, which most conferencing products do not enable by default.
In the Ethora ecosystem
Ethora calls run on WebRTC, so every media leg between a participant and the SFU is carried over DTLS-SRTP as the standard requires; there is no mode in which call audio or video crosses the network unencrypted. Chat and API traffic use TLS in transit, and connectivity for calls follows the standard ICE model described under STUN and TURN.
Ethora does not add a client-side end-to-end layer on top of SRTP. That is the same deliberate choice the platform makes for messaging: the server is trusted so that regulated customers can archive, audit and moderate, and the deployment model (managed cloud, dedicated server in your own cloud account, or self-hosted) decides who operates that server. If your threat model requires that no server can ever access media, read the end-to-end encryption entry before choosing a platform.