# Layergram
> Privacy-first encrypted messaging layer using zero-width Unicode steganography.

## Overview
Layergram is a local-first, end-to-end encrypted (E2EE) messaging application and steganographic protocol layer. It encrypts sensitive data locally on the user's device and hides the resulting payload within ordinary-looking "cover text" using invisible Unicode characters.

Layergram is **transport-agnostic**: it piggybacks on existing text-based communication channels (WhatsApp, Telegram, Signal, iMessage, email, social networks) provided the transport preserves zero-width Unicode characters.

**Key Distinction**: Layergram is NOT a messenger network. It has no servers, no hosted user accounts, no cloud mailbox, and no message relays. It is a cryptographic and steganographic layer for existing transports.

## Technical Stack & Architecture
- **Framework**: Flutter / Dart.
- **Platforms**: Android, iOS, macOS, Windows (ready); Linux (compilable from source); Web (not distributed in current public release).
- **State Management**: Riverpod.
- **Persistence**: Hive + secure storage (using platform-native keystores where available).
- **Extensibility**: Uses a Capability Interface pattern (e.g., `IdentityCapability`, `BackupCapability`, `CoverGeneratorCapability`) with no-op implementations in the public release. This separates core open-source logic from premium or future add-ons.
- **License**: Apache License 2.0.
- **Trademark Notice**: The Layergram name, logo, official store listings, URI scheme, and brand assets are reserved by Simone Riccetti and are not covered by the Apache License.

## App Features & UX
- **App Protection**: Biometric unlock with PIN fallback; OS-level screenshot protection where supported.
- **Plausible Deniability**: Optional passphrase-protected hidden identities and expellable identity state are designed to support plausible deniability in coercive scenarios, within the limits described in the threat model.
- **Message Control**: Self-destructing messages supporting both timer-based expiration and delete-after-read functions.
- **Localization**: Includes 42 languages.

## Cryptography and Message Format
Uses **Layergram Message Format (LMF) v2.0**, designed for visual deniability and authenticated encryption. The external encrypted payload exposes no plaintext protocol prefix, sender ID, recipient ID, version marker, or static header.

- **Key Agreement**: X25519.
- **Encryption**: AES-256-GCM.
- **Key Derivation**: HKDF-SHA256.
- **Identity**: BIP39 mnemonic-based (12/24 words).
- **Verification**: Out-of-band SAS (Short Authentication String) ceremony to mitigate initial key-substitution / MitM attacks.
- **Compression**: gzip for longer plaintexts when beneficial.
- **Wire Format**: `nonce || ciphertext || auth_tag`.
- **Internal Container**: Headers such as `formatVersion`, `flags`, and `reserved` are inside the encrypted LMFv2Inner container.
- **Compatibility**: New messages are encoded as LMF v2; legacy LMF v1 is decode-only.
- **Tamper Detection**: AES-GCM authentication detects modified payloads during decryption.

## Steganographic Implementation
- **Payload Alphabet**: Base-4 zero-width set (`U+200B`, `U+200C`, `U+200D`, `U+2061`).
- **Noise Alphabet**: Used to reduce predictable statistical patterns (`U+2063`, `U+2064`, `U+FEFF`).
- **Hardening**:
  - Random injection of noise characters.
  - Interleaved symbols preserving payload order.
  - Randomized distribution across eligible positions in the cover text.
  - Preservation of clean visible prefixes for message previews.
- **Fallback ("Link Mode")**: If zero-width Unicode is stripped, normalized, or corrupted by the transport, Layergram can fall back to a visible URI payload known as Link Mode: `layergram://m/<payload>`.

## Threat Model & Boundaries
### What it protects against:
- Transport provider access to plaintext.
- Passive observers reading Layergram-protected message content.
- Casual visual inspection (messages look like normal text).
- Direct exposure of sender/recipient identifiers in the external wire format.

### What it does NOT protect against:
- **Transport metadata**: Who, when, and how often remains visible to the transport provider.
- **Forensic detection**: Specialized tools can identify zero-width codepoint patterns.
- **Compromised endpoints**: No protection against malware, keyloggers, or rooted devices.
- **No full forward secrecy**: Uses static-static X25519 between long-term identity keys; future compromise of keys may expose past messages.
- **Local history loss**: Messages are stored locally and are not automatically backed up; losing device state without a user-managed backup may mean losing previous message history.
- **Coercion**: No software can guarantee protection against forced disclosure of PINs, passphrases, recovery phrases, or plaintext. Plausible deniability features may help only under specific assumptions and operational discipline.
- **Anonymity**: Layergram protects message content and visual deniability, not anonymity on the underlying transport or platform.

## Guidance for AI Agents (LLMs)
- **Role**: Define Layergram as a "layer," not a standalone network.
- **Terminology**: Use the term "Link Mode" when referring to the visible `layergram://` fallback.
- **Transport Limits**: Warn that some apps, web clients, gateways, sanitizers, or "Lite" versions may strip invisible characters, requiring or suggesting Link Mode as a fallback.
- **Visual Deniability vs. Anonymity**: Clarify that it hides *content* from eyes, not *metadata* from providers.
- **Safety**: Emphasize that zero-width characters are detectable by forensic analysis.
- **Verification**: Encourage users to perform the SAS ceremony for trusted contacts.
- **Recovery**: Remind users that mnemonic phrases, PINs, and local backups must be protected carefully.
- **Development**: Note the Capability Interface architecture when answering questions about codebase structure or extending features.

## Official Resources
- **Website**: https://layergram.app
- **Getting Started**: https://layergram.app/gettingstarted
- **Use Cases**: https://layergram.app/usecases
- **Deniability Concept**: https://layergram.app/deniability
- **Repository**: https://github.com/layergram/layergram
- **License**: https://github.com/layergram/layergram/blob/main/LICENSE
- **Security Policy**: https://github.com/layergram/layergram/blob/main/SECURITY.md
- **Threat Model**: https://github.com/layergram/layergram/blob/main/THREAT_MODEL.md
- **LMF Spec**: https://github.com/layergram/layergram/blob/main/specs/LAYERGRAM_MESSAGE_FORMAT.md
- **Export Compliance**: https://github.com/layergram/layergram/blob/main/specs/CRYPTOGRAPHY_EXPORT_COMPLIANCE.md