avatar
zerokn0wledge 🪬✨
25,013 Twitter followers
Follow
terminally onchain | co-founder @RedactedRes | cooking @steak_studio | vibe coder | hyperliquid maxi | biττensor believer
Posts
avatar
zerokn0wledge 🪬✨
Zama’s FHE stack overview FHE lets you compute on encrypted data without decryption. Input stays encrypted, operations run on ciphertext, output comes back encrypted. Only data owner can decrypt results. @zama built a production FHE stack with 3 layers: 1. TFHE-rs (core library) Implements TFHE scheme (Torus Fully Homomorphic Encryption) in Rust. Handles encrypted integer and boolean operations. For context, addition takes 1ms, multiplication 20ms on encrypted 8-bit integers 2. Concrete (compiler) converts Python/Rust code to FHE circuits automatically. Developer writes normal functions, compiler transforms them to work on encrypted data. Optimization layer reduces circuit depth - each operation adds noise to ciphertext. After 1000 operations, noise exceeds threshold and requires bootstrapping (noise refresh that takes 50-150ms per operation). 3. fhEVM (blockchain integration) - EVM-compatible smart contracts processing encrypted inputs. Deployed on Zama's devnet with 8-second block times. Gas costs run 100-1,000x higher than standard EVM operations due to FHE computation overhead. Zama processed 2.4M encrypted transactions in 2024 testnet and 6.7M transactions to date. Use cases include blind auctions (bids stay encrypted until reveal), private DeFi (balance/trade amounts hidden), confidential voting. Only works for specific operation types. Neural network inference possible but 10,000x slower than plaintext. Best for low-complexity operations on sensitive data where confidentiality justifies performance penalty.
ZAMA
17.09%
loading indicator
Loading..