Coverage and scope
Sogool protects one-to-one secure chat messages and in-chat file content with AES-256-GCM. Calls use two separate encryption layers: Sogool application-level media-frame encryption and WebRTC DTLS-SRTP transport encryption. Peer-to-peer transport is not itself the content encryption algorithm.
| Data path | Content protection | Scope / condition |
|---|---|---|
| Private 1:1 chat | AES-256-GCM | Extra context verification in supported, negotiated sessions |
| In-chat photos and files | AES-256-GCM | File-scoped material; chunk/hash checks when negotiated |
| Offer / Answer / ICE signaling | AES-256-GCM | After session DH agreement |
| Video frames | App AES-256-GCM + DTLS-SRTP | Reviewed native video-call path |
| Audio within video calls | App AES-128-GCM + DTLS-SRTP | Reviewed native video-call path |
| Audio-only calls | App AES-128-GCM + DTLS-SRTP | Separate compatibility KDF path |
Outside this review: public/multi-party rooms, standalone encrypted notes/files, and their local storage or backup mechanism are not established by this 1:1 communication profile.
Invitation, contacts, and connection context
A user may establish a known contact through invitation/friend flows or separately share a 16-character hexadecimal connection context. Users are not required to enter an encryption key for each call.
- The 16-character connection context is not the AES-256 encryption key.
- Invitation or context information should reach the intended person over a trusted channel.
- The contact selection mechanism and cryptographic identity verification are related but distinct.
Identity authentication and TOFU
Sogool's described design pins the contact's long-term ECDSA P-256 identity key on first trust, then compares it on subsequent connections. Session DH public keys are validated with ECDSA P-256/SHA-256 signatures under that identity.
- Trust On First Use (TOFU): later key changes can be detected; a malicious first encounter is not automatically excluded.
- Ordinary pinned-key mismatches are rejected by design.
- There is a constrained authenticated recovery / identity-key succession exception for existing contacts; this is not unrestricted key replacement.
An external evaluator should test first-contact substitution, later mismatches, and the constrained recovery path separately.
Session agreement and key separation
The described session uses RFC 3526 Group 15 3072-bit MODP Diffie–Hellman with signed session public keys. The resulting shared value is processed by HKDF-SHA-256, yielding purpose-separated key material.
| Usage | Key-material domain |
|---|---|
| Private messages | CHAT |
| In-chat files | FILE |
| Signaling | SIGNALING |
| Call audio | CALL-AUDIO |
| Call video | CALL-VIDEO |
Purpose separation reduces accidental key reuse across communication domains. HKDF alone does not prove peer authentication or forward secrecy.
Protected signaling and connection setup
WebRTC setup exchanges Offer, Answer and ICE material. The published security profile describes this material being protected with AES-256-GCM using signaling-specific material after DH agreement.
- Exchange connection context and signed session key material.
- Authenticate the session key and derive the shared secret.
- Encrypt and authenticate Offer / Answer / ICE data.
- Do not apply unauthenticated signaling content as valid connection settings.
- Establish WebRTC, then exchange real-time content directly between devices.
Invitation, signaling and notifications may still use service infrastructure. This is different from a server storing or relaying message bodies.
Message encryption, context binding and replay
Private chat payloads are protected with AES-256-GCM, which provides both confidentiality and authentication. For 1:1 sessions where both peers support and negotiate the relevant feature, additional per-message derivation and context checks are described.
- Session context, type, sender/recipient direction, sequence, unique message ID and time are combined with current text material and HKDF-SHA-256 to derive 256-bit per-message key material.
- This material is then passed to the AES module; the HKDF output must not automatically be described as the final raw AES key.
- Authenticated interior context is compared with external metadata; not all external metadata is confidential.
- Message ID / sequence receive state is committed after authentication and context validation, supporting replay checks.
Per-message derivation, replay checks and key refresh are distinct controls and are not guaranteed on every legacy session.
In-session text-key refresh
In feature-negotiated 1:1 sessions, a new ECDH P-256 agreement may be initiated according to time/message-count conditions. The procedure authenticates fresh public keys using ECDSA P-256/SHA-256, performs HMAC-SHA-256 key confirmation, then combines old text context and new shared material with HKDF-SHA-256.
- The feature refreshes text-session material, not file, signaling and media keys together.
- An incomplete refresh is not designed to trigger plaintext chat fallback.
- This must not be equated with Signal's Double Ratchet or a universal forward-secrecy guarantee.
In-chat file encryption and chunk integrity
In-chat photos and files are intended for encrypted device-to-device transfer, rather than upload to a storage server followed by download. File-specific key material is derived from the file domain, an identifier and randomness, then used with AES-256-GCM.
Where both peers negotiate the relevant verification feature:
- Authenticated control data binds ciphertext SHA-256, size and file ID.
- The receiver checks file membership, chunk index and chunk count.
- A final whole-ciphertext hash and size is verified.
- Missing, substituted, reordered or mismatching chunks do not constitute a valid completed file.
A raw SHA-256 digest is not authentication on its own; control-message authentication, GCM authentication and file structure checks work together.
Two-layer encryption for calls
Call protection comprises two independent layers: (1) Sogool encrypts encoded media frames at the application layer, then (2) WebRTC carries those already-encrypted frames under DTLS-SRTP. The receiver reverses these steps and authenticates content before playback.
| Media type | Layer 1: Sogool frame | Layer 2: transport |
|---|---|---|
| Video | AES-256-GCM | WebRTC DTLS-SRTP |
| Audio within video calls | AES-128-GCM | WebRTC DTLS-SRTP |
| Audio-only call | AES-128-GCM | WebRTC DTLS-SRTP |
Do not describe all media as AES-256. The documented audio frame algorithm is AES-128-GCM. Nor does the static review prove both layers are active on every deployed Android/iOS pairing.
Native media processing is separate from the web UI. “Native” is not, by itself, a cryptographic security rating.
Media key schedule, IV, tag and AAD
Call key material is separated by call, direction, and media purpose. The video-call final-frame KDF is HKDF-SHA-256; the existing native audio-only compatibility path retains a separate PBKDF2 profile. Therefore the same final KDF must not be claimed for every call path.
The reviewed video-call frame implementation is described with:
- a 96-bit random IV and 128-bit GCM authentication tag;
- necessary clear codec header bytes authenticated as AAD (additional authenticated data);
- a finite recent-receive IV window used to detect replay of previously authenticated frames;
- rejection of unauthenticated frame content as valid media.
A finite IV window does not prevent replay of every frame for all time. These specific frame-layout details should not be projected onto all legacy audio-only implementations without review.
WebRTC P2P, STUN and the role of servers
Once ICE establishes the connection, real-time chat, files and calls use direct device-to-device WebRTC transport. The published profile states TURN relaying is disabled, while STUN assists NAT/path discovery.
- A direct connection may fail on restrictive networks when TURN is not available.
- Invitations, connection coordination and push notifications can use separate service infrastructure.
- P2P is not anonymity: IP addresses, connection times or traffic volume may be visible to relevant participants or networks.
- A design that does not store or relay message bodies must not be described as having no servers whatsoever.
Testing the actual selected ICE candidate path is a separate runtime verification step.
Authentication failure, compatibility and downgrade
The described design rejects unauthenticated content rather than rendering it as valid and does not deliberately substitute an unencrypted call when app-level frame encryption cannot be prepared.
- Legacy peers may use an older encrypted compatibility format.
- Message-context validation, per-message derivation and text-key refresh require mutual feature negotiation.
- A previously pinned or negotiated higher-security feature is protected against unauthorized downgrade by separate checks.
- A rejected media frame does not necessarily terminate the whole call.
Reviewers should distinguish an allowed older encrypted protocol from a malicious downgrade of an established security level.
Privacy and security limitations
Encryption and direct transport do not eliminate every threat:
- incorrect first-contact identification or compromised invitation channels;
- compromised endpoints, screen capture, recipient recording and key theft;
- possible exposure of network metadata such as IP addresses and timing;
- the fact that some externally visible metadata is authenticated but not confidential;
- unreviewed public rooms, standalone local secure notes/files, backup and recovery behaviors.
The term AES-256 by itself cannot establish an application-wide security level. Key generation, persistence, runtime paths, platform code and deployed binaries also matter.
Evidence, validation gaps and test matrix
This page summarizes the published CHAT.zip security profile, which describes a targeted static comparison of related JavaScript paths and native video-frame code. It is not an independent third-party audit, a whole AES-module audit, a deployed-binary inspection, or a completed attack test.
| Area | Documented static design | Validation still required |
|---|---|---|
| Identity/DH | Signed keys, TOFU, MODP-3072, HKDF | First-contact substitution, mismatches and recovery exception |
| Messages | AES-GCM, contextual material, receive sequence | Replay, tampering and failure-state behavior |
| Text refresh | ECDH, signature, HMAC, HKDF | Delays, duplication, expiry and compatibility |
| Files | Per-file material, authenticated hashes, chunk checks | Missing/reordered/mismatched chunks and final hash |
| Calls | Native frame cryptor and DTLS-SRTP design | Android↔Android, iOS↔iOS and cross-platform live tests |
| Implementation | Public reviewed profile | Full AES module, KDF compatibility/defaults and fallback paths; actual releases |
The documented static design is not a pass/fail certificate. Reproduction tests must also verify lack of plaintext fallback after authentication or encryption failure.
Primary documents, standards and download
Download original CHAT.zip technical pack — includes README.md, SECURITY_PROFILE.json, SECURITY_REVIEW_SCOPE.md, STANDARDS.md, CHAT_SECURITY_PROFILE.js and SHA256SUMS.txt. This is public documentation, not the production app code, cryptographic secrets or deployed binaries.
- AES-GCM: NIST SP 800-38D
- HKDF-SHA-256: IETF RFC 5869
- 3072-bit MODP DH: IETF RFC 3526, Group 15
- WebRTC security: IETF RFC 8827
- WebRTC transport: IETF RFC 8835
The referenced standards do not certify the complete Sogool protocol. If this documentation differs from a deployed app version, verified production behavior takes precedence.
Source: /src/CHAT.zip · This is static HTML so crawlers and analysis tools can access the complete technical text without JavaScript.