We present a lightweight, practical protocol for metadata privacy and strong encryption simultaneously in instant messaging, which we call Deniable Instant Messaging (DenIM). Our main insight is that by proving privacy on a per-application basis using information flow techniques, we can reuse infrastructure which allows us to share the same servers between sensitive and non-sensitive communication. Additionally, our solution is also censorship resilient as we adhere to the Cute Cat Theory of Censorship by reusing infrastructure.
DenIM is designed around an existing, commonly used instant messaging protocol with strong security guarantees (Signal), and is intended to be deployed in real instant messaging platforms (e.g., Signal, WhatsApp, Facebook Messenger; all already use the Signal protocol). One of our observations is that to provide privacy to any one person, we must provide privacy to everyone to be able to offer strong guarantees—the hide-in-the-crowd-effect only works when everyone is in on it. This observation means privacy needs to be available in everyday tools (e.g., instant messaging) instead of just in single-purpose privacy tools (e.g., Tor).
