What Is the Signal Protocol?
The Signal Protocol is the end-to-end encryption protocol used by the Signal messenger, WhatsApp, and a growing number of secure communication platforms, including Sanket. It was designed by Trevor Perrin and Moxie Marlinspike and has been studied more closely than almost any other messaging protocol, including a formal academic security analysis published in 2017.
In one sentence: the Signal Protocol lets two devices agree on secret keys over an untrusted network and then change those keys with every message, so that only the sender and the recipient can ever read what was sent - not the server, not the network operator, and not the platform provider.
It is built from three main parts:
- X3DH key agreement - how two devices establish a shared secret, even when one of them is offline.
- The Double Ratchet - how that secret evolves into a new key for every single message.
- Sender Keys - how the same guarantees are extended efficiently to group chats.
Step 1: Starting a Conversation with X3DH
When you register a Sanket device, it generates several key pairs locally:
- A long-term identity key, which represents the device.
- A signed pre-key, rotated periodically and signed by the identity key.
- A batch of one-time pre-keys, each used once and then discarded.
Only the public halves are uploaded to the server. The private keys never leave the device.
When someone wants to message you for the first time, their device fetches your public keys from the server and combines them with its own keys through several Diffie-Hellman calculations. This is the Extended Triple Diffie-Hellman (X3DH) handshake. Both devices arrive at the same shared secret without that secret ever crossing the network, and it works even if you are offline when the first message is sent.
Signal has since published PQXDH, a successor to X3DH that adds a post-quantum Kyber key exchange to the same handshake, and SANKET uses it. The design principle is unchanged: the server distributes public keys, but it never learns the shared secret.
Step 2: A New Key for Every Message with the Double Ratchet
Agreeing on one secret is not enough. If the same key protected an entire conversation, stealing it once would expose everything. The Double Ratchet solves this by deriving a fresh key for every message.
It combines two ratchets:
- The symmetric-key ratchet. Each message key is derived from a chain key, and the chain key then moves forward through a one-way function. You can compute the next key from the current one, but never go back. Once a message key is used, it is deleted.
- The Diffie-Hellman ratchet. Every time the conversation changes direction, the devices exchange new Diffie-Hellman public keys and mix fresh randomness into the chain. This resets the chain with secrets an attacker has never seen.
The result is two properties that matter a great deal for sensitive communication:
| Property | What it means in practice |
|---|---|
| Forward secrecy | If an attacker steals a device's current keys, they still cannot decrypt messages sent before the theft. Old message keys no longer exist. |
| Post-compromise security | After the next exchange of new Diffie-Hellman keys, the attacker is locked out again, even if they keep a copy of the stolen keys. |
Inside each step, libsignal encrypts the message content with AES-256 and authenticates it with HMAC-SHA256, so a tampered message is rejected instead of decrypted.
Step 3: Groups with Sender Keys
Encrypting a separate copy of every message for every group member would be slow in large groups. The Sender Keys scheme solves this. Each member creates a sender key for the group and shares it with every other member through their individual, already encrypted Signal sessions. Group messages are then encrypted once with the sender key and delivered to everyone.
When someone leaves a group, keys are rotated so that the departed member cannot read new messages. For organisations, this is the cryptographic side of something administrators care about: removing a person from a group must actually remove their access.
Safety Numbers: Verifying Who You Are Talking To
End-to-end encryption only helps if you are encrypting to the right person. Every Signal Protocol conversation has a safety number derived from both identity keys. Two users can compare it in person or over a trusted channel. If someone's identity key changes, for example because they reinstalled the app or because someone is attempting interception, the app warns you before you continue.
What the Signal Protocol Does Not Protect
A protocol is not a complete security programme. Being honest about the limits is part of evaluating any secure messenger:
- Metadata. The protocol protects message content. The server still needs some routing information to deliver messages. How much metadata exists, where it lives, and who can access it depends on the platform's architecture and deployment, not the protocol.
- Compromised endpoints. If a device is unlocked and in an attacker's hands, or infected with spyware, the attacker can read what the user reads. Device security, screen lock, and remote revocation matter.
- Poor implementations. A correct protocol can be undermined by a homegrown implementation, a weak random number generator, or a "fallback" mode that quietly skips encryption.
- Who runs the server. End-to-end encryption hides content from the server, but the server operator still controls availability, account provisioning, and stored metadata. Jurisdiction and hosting still matter.
How Sanket Uses the Signal Protocol
Sanket was designed around these limits rather than around the protocol alone.
Official libsignal only. Sanket uses Signal's own audited library, libsignal, on mobile and desktop. There is no custom cryptography, no homegrown reimplementation, and no fallback mode. If the official library is not available on a platform, the feature does not ship there.
Keys generated and kept on the device. Identity keys and pre-keys are generated on the device and stored in hardware-backed secure storage - the iOS Keychain and the Android Keystore. They are never uploaded and never included in backups.
A zero-knowledge server. The Sanket server relays and stores encrypted blobs. A full compromise of the server reveals ciphertext, not conversations.
Post-quantum protection. New sessions use PQXDH, and the Double Ratchet runs alongside SPQR, Signal's post-quantum companion ratchet. See post-quantum encryption.
AES-256 throughout. Messages use libsignal's AES-256 encryption. Every file has its own AES-256-GCM key, and calls use AES-256-GCM frame encryption. See encryption in SANKET.
Your infrastructure, your jurisdiction. Sanket can be deployed in your own data centre, on a national cloud, or in a fully air-gapped network. This addresses the parts of the threat model that encryption alone cannot: who operates the server, where metadata lives, and which country's laws apply to it.
Governance around the cryptography. Administrators provision users, enforce device policy, and revoke access instantly. Lost devices can be revoked so they no longer receive messages.
Signal Protocol vs. Server-Side Encryption
Many enterprise collaboration tools describe themselves as "encrypted". Usually that means encryption in transit (TLS) and at rest, with the provider holding the keys. The difference is fundamental:
| Server-side encryption | Signal Protocol end-to-end encryption | |
|---|---|---|
| Who can decrypt messages | The provider, and anyone who compels or compromises it | Only the sender's and recipients' devices |
| Effect of a server breach | Content can be exposed | Attacker gets ciphertext only |
| Effect of a stolen key | Potentially all data protected by that key | Only the messages that key protected, then the ratchet moves on |
| Lawful access requests to the provider | Provider can hand over content | Provider has no content to hand over |
This is why, when organisations compare a Microsoft Teams alternative or a WhatsApp alternative for business, the encryption model should be one of the first questions, not a footnote.
Frequently Asked Questions
Is the Signal Protocol the same as the Signal app? No. The Signal app is one messenger that uses the protocol. The protocol itself is used by several products, including WhatsApp and Sanket. What differs between products is the implementation, the server architecture, who controls the deployment, and what happens to metadata.
Has the Signal Protocol been independently analysed? Yes. It has been the subject of extensive academic research, including formal security proofs of its core properties. Sanket relies on Signal's official libsignal library rather than a reimplementation, so it benefits from that scrutiny directly.
Can quantum computers break the Signal Protocol? The concern is that traffic recorded today could be decrypted by a future quantum computer. Signal has responded with PQXDH, which adds a post-quantum key exchange to the initial handshake, and SPQR, a post-quantum ratchet. SANKET uses both. Symmetric encryption with 256-bit keys, such as AES-256, is considered resistant to known quantum attacks.
Does end-to-end encryption make a platform compliant with GDPR, NIS2, or defence security requirements on its own? No. Encryption is a necessary control, but compliance also depends on hosting, jurisdiction, access governance, audit logging, and incident response. See our buyer's guide to secure communication platforms for the full checklist.
