FAQ

What is wacrg?

The WhatsApp Calls Research Group: a collaborative, GitHub-native effort to reverse-engineer and document a complete, machine-readable specification of the WhatsApp 1:1 call protocol, covering signaling, keying, media, and transport. It is maintained by people who work on Baileys and unifies findings from multiple reverse-engineering techniques into one provenance-tracked spec. Start at the home page.

Is this an official WhatsApp project?

No. wacrg is not affiliated with, authorized by, or endorsed by WhatsApp or Meta. It is independent interoperability and security research. See legal and ethics and DISCLAIMER.md.

Is the spec accurate / complete?

Not yet, and we are upfront about it. As of 0.1.0 this is a research scaffold with a starting model, not authoritative truth. There are no real captures; all examples are synthetic. Most facts are graded probable or speculative and carry open questions. Read each fact's confidence and provenance before relying on it.

Why does every fact have a "confidence" and "provenance"?

Because independent contributors use different techniques, and we need a way to converge honestly. Provenance records who saw a fact and where; confidence records how much to trust it. A fact is promoted toward confirmed only when independent techniques corroborate it. This corroboration rule is the heart of the project.

What are the confidence levels?

confirmed, probable, speculative, unknown. Most of the current spec sits at probable (signaling) or speculative (media/keying derivation).

What are the seven techniques?

websocket-capture, baileys-instrumentation, frida-hooking, mitm-tls, static-smali-analysis, memory-dump, and wasm-analysis. They reveal different layers; the techniques page details each one's strengths and limits.

How do I contribute a finding?

File a Stanza capture issue using the GitHub Issue Form. A workflow ingests it into corpus/captures/ and opens a PR automatically. See the capture pipeline. You can also add a capture by hand by copying the example.

Can I capture my own real calls and submit them?

Only with synthetic test accounts you control, and only after full sanitization (no PII, no key material). Never capture other people. Never paste real bytes. See the sanitization rules. This is non-negotiable.

Why is the media plane so under-documented?

Because the cheap techniques (WebSocket capture, Baileys instrumentation) only see the signaling plane; the media plane is SRTP/UDP and never crosses the control socket. Observing it needs the harder techniques (Frida, memory dump) or WASM analysis of the WhatsApp Web client, whose calling engine ships as a WebAssembly module, so it is the project's biggest open frontier. See the roadmap.

What is the difference between corpus/ and spec/?

corpus/ is raw intake: individual captures as observed, with provenance. spec/ is the curated source of truth: captures reconciled into stanzas, flows, enums, and techniques with confidence levels. Human docs under docs/spec/ are generated from spec/ and must not be edited by hand.

Why is the documentation generated from YAML?

So there is a single machine-readable source of truth. Validation, generation, and the coverage metric all run off the same YAML, which keeps the human docs, diagrams, and stats consistent and lets us automate quality checks. Edit the YAML under spec/, not the generated Markdown.

What does the coverage percentage mean?

A weighted measure of how much of the spec is well-supported: (confirmed + 0.5 × probable) / total, broken down by category and technique. We use it to find gaps and aim the next captures. See coverage and the roadmap.

Does this cover group calls?

No. wacrg is scoped to 1:1 calls. Group-call media topology is explicitly out of scope for now.

I found a security vulnerability. What do I do?

Do not post it as a capture. Follow SECURITY.md and report it responsibly first.

What license is this under?

Code/tooling: MIT. Spec and docs content: CC BY 4.0. See legal and ethics.