Voice & Video Calls

Technical specification of real-time encrypted communication.

Overview

Zentalk uses WebRTC for peer-to-peer voice and video calls, with DTLS-SRTP for media encryption and E2EE for signaling.

Caller

Signaling

Callee

SDP Offer

ICE

SDP Answer

DTLS Handshake

ClientHello

ServerHello

Certificate

Finished

SRTP Encrypted Media

Audio (AES-GCM)

Video (AES-GCM)

Perfect Forward Secrecy: Keys derived per-session, never transmitted

Protocol Stack

Layer	Protocol	Purpose
Signaling	E2EE (Double Ratchet)	SDP exchange, ICE candidates
Key Exchange	DTLS 1.2	Derive SRTP keys
Media Transport	SRTP	Encrypted audio/video
NAT Traversal	ICE/STUN/TURN	Connection establishment

Signaling (E2EE)

Call setup messages are encrypted using existing Double Ratchet sessions:

Message	Content	Encryption
Call Offer	SDP, DTLS fingerprint	Double Ratchet
Call Answer	SDP, DTLS fingerprint	Double Ratchet
ICE Candidates	Address:port pairs	Double Ratchet
Call End	Termination signal	Double Ratchet

SDP Security


Encrypted SDP contains:
- Media codecs (Opus, VP8/VP9)
- DTLS fingerprint (SHA-256)
- ICE credentials
- Candidate addresses

The server never sees SDP content - only encrypted blobs.

DTLS-SRTP

Datagram Transport Layer Security establishes media encryption keys:

Handshake

Step	Message	Purpose
1	ClientHello	Cipher suite proposal
2	ServerHello	Cipher selection
3	Certificate	Self-signed, fingerprint verified
4	KeyExchange	ECDHE key exchange
5	Finished	Handshake complete

Fingerprint Verification


1. Alice includes DTLS fingerprint in SDP (E2EE signaling)
2. Bob receives fingerprint via Double Ratchet
3. During DTLS handshake, Bob verifies certificate matches fingerprint
4. If mismatch → terminate call (MITM detected)

Key Derivation


DTLS Master Secret
    │
    ├─► SRTP Encryption Key (AES-128)
    ├─► SRTP Authentication Key (HMAC-SHA1)
    └─► SRTP Salt (112 bits)

SRTP Parameters

Parameter	Value
Cipher	AES-128-GCM or AES-128-CTR
Authentication	HMAC-SHA1-80 (if not GCM)
Key size	128 bits
Salt	112 bits
Replay protection	64-bit sequence window

Packet Format


┌─────────────────────────────────────────┐
│ RTP Header (12 bytes)                   │
├─────────────────────────────────────────┤
│ Encrypted Payload (variable)            │
├─────────────────────────────────────────┤
│ Authentication Tag (10 bytes)           │
└─────────────────────────────────────────┘

ICE (Interactive Connectivity Establishment)

Candidate Types

Type	Description	Privacy
Host	Local IP address	Reveals LAN IP
Server Reflexive	Public IP via STUN	Reveals public IP
Relay	TURN server relay	Hides both IPs

TURN Servers

For maximum privacy, relay candidates are preferred:

Property	Value
Protocol	TURNS (TLS encrypted)
Authentication	Ephemeral credentials
Credential lifetime	24 hours
Server selection	Geographic proximity

Connection Priority


1. Direct peer-to-peer (if possible)
2. STUN-assisted (NAT traversal)
3. TURN relay (fallback)

Audio Encryption

Opus Codec

Parameter	Value
Sample rate	48 kHz
Bit rate	6-510 kbps (VBR)
Frame size	20 ms
Channels	1 (mono) or 2 (stereo)

Per-Packet Encryption


For each audio packet:
  1. Encode audio → Opus frame
  2. Encrypt with SRTP key
  3. Add authentication tag
  4. Transmit via UDP

Video Encryption

Codecs

Codec	Usage
VP8	Default, good compatibility
VP9	Higher efficiency
H.264	Hardware acceleration

Selective Forwarding (SFU)

For group calls, video uses SFU architecture:

Property	Value
Encryption	E2EE (insertable streams)
SFU visibility	Encrypted bitstream only
Key distribution	Via E2EE signaling

Group Calls

Architecture

Participants	Method
2	Peer-to-peer
3-6	SFU (Selective Forwarding)

Group Call Keys


1. Initiator generates Media Key (MK)
2. MK distributed via Double Ratchet to each participant
3. All participants use MK for SRTP
4. Key rotation on participant change

Insertable Streams API


Encryption Pipeline:
  Camera → Encode → Encrypt(MK) → SFU → Decrypt(MK) → Decode → Display

SFU sees: Encrypted bitstream (cannot decode)

Security Properties

Property	Mechanism
Confidentiality	SRTP (AES-128-GCM)
Integrity	GCM tag / HMAC-SHA1
Authenticity	DTLS fingerprint verification
Forward Secrecy	ECDHE in DTLS
Replay Protection	SRTP sequence numbers

Metadata Protection

Data	Protection
Call initiation	3-hop relay
Call duration	Local only, not logged
Participant IPs	TURN relay option
SDP content	E2EE encrypted

Call States

State	Description
IDLE	No active call
OFFERING	SDP offer sent
ANSWERING	SDP answer sent
CONNECTING	ICE/DTLS in progress
CONNECTED	Media flowing
DISCONNECTED	Call ended

Error Handling

Error	Action
DTLS fingerprint mismatch	Terminate, warn user
ICE timeout	Fallback to TURN
SRTP auth failure	Drop packet
Connection lost	Attempt reconnect (30s)

Protocol Specification - E2EE for signaling
Architecture - System components
Features - Feature overview