Messages in a coded chat are encrypted with a key derived from the chat's 3-character code (details in PROTOCOL.md). The goal is narrow: someone nearby, running this app or anything that speaks the protocol, shouldn't be able to read your messages just by pointing a camera at your screen. The Open channel has no code and no encryption; anyone in view can read it.

• A determined attacker with a recording. There are only 46,656 possible codes. Key derivation is deliberately slow (PBKDF2-SHA256, 50,000 iterations), but anyone who records a message can still try every code offline. Our estimate is under a minute on a laptop and seconds on a GPU. The code keeps out casual onlookers; it is not real secrecy.
• Forgery or tampering. Messages are not authenticated (AES-CTR with a CRC-16, no MAC). Someone who can shine their own screen at your camera could send you messages, or alter a message they captured, without knowing the code. The app can't prove who sent a message.
• Seeing the screen itself. The sender types in plain view. Encryption covers the flashing light, not the keyboard.
• The delivery confirmation. The flashlight blink that tells the sender "received" is not tied to a message or a code. Anyone with a flashlight can fake it.

Messages and chat codes are stored unencrypted in the app's private storage on the phone (see PRIVACY.md), protected by iOS's normal app sandbox and device encryption.

If you need confidentiality against a motivated adversary, use something else.

Please use GitHub's private reporting (Security › Report a vulnerability on this repository) rather than a public issue. This is a hobby project with no bounty, but reports are read and fixed.