// case study — 02/05

CipherMind

End-to-end encrypted AI chat

CipherMind screenshot

overview

Encrypted AI chat where every message is AES-256-GCM ciphered before leaving the browser — zero plaintext in transit. Session keys are derived via PBKDF2 and integrity is guaranteed by HMAC-SHA512. The Groq Llama 3.3 70B backend runs server-side with the API key fully protected. Tracks encrypted/decrypted message counts, token usage, and HMAC checks per session.

metrics

AES-256
GCM encryption
zero
plaintext in transit
Llama 3.3 70B
Groq backend

architecture

  1. 01PBKDF2 — session key derivation
  2. 02AES-256-GCM — message encryption in-browser
  3. 03HMAC-SHA512 — integrity verification
  4. 04Groq API — Llama 3.3 70B server-side
  5. 05Session store — encrypted history + export

stack

  • React
  • TypeScript
  • AES-256-GCM
  • PBKDF2
  • HMAC-SHA512
  • Groq API

links

github ↗live site ↗

next project

ChainAudit