Privacy tech toolbox: TEE, MPC, FHE and ZK trade-offs
Thread179 words41 likes
Lays out the distinct trust and cost trade-offs of TEE, MPC, FHE and ZK, stressing they are not interchangeable and that real systems combine them.
“Privacy tech” often gets treated like a single checkbox, but it’s really a toolbox with very different tradeoffs.
➤ TEE (Trusted Execution Environments): runs code in a hardware “enclave.” You get near-native speed, but you’re trusting specific hardware + its attestation + its patch story.
➤ MPC (Multi-Party Computation): splits a computation across multiple parties so no single party sees the full input. Trust shifts from “one secure box” to “many parties don’t collude,” but the cost is coordination and extra compute.
➤ FHE (Fully Homomorphic Encryption): compute directly on encrypted data. The trust model is attractive (no party needs to see plaintext), but today it’s still expensive and use-case constrained.
➤ ZK (Zero-Knowledge proofs): prove a statement about data without revealing the data. Great for verification and selective disclosure, but proving costs (and circuit complexity) shape what’s practical.
The key point: these aren’t interchangeable. Many real systems combine them (e.g., TEEs to speed up parts, ZK to verify results, MPC for distributed custody).
Also linking our @a1research__ report from a while ago that covered some of these concepts:
1 image on X