Privacy in crypto has always been a trilemma: you can have two of three — privacy, composability, and performance. But historically, not all three at once.
Privacy without composability: Zcash. You can shield transactions, but shielded funds can't easily interact with other DeFi protocols. Privacy is a dead end.
Composability without privacy: Ethereum/Solana. Perfectly composable DeFi ecosystem. Zero privacy. Every trade visible.
Privacy without performance: Early FHE systems. Mathematically private, but 100-1000x slower than plaintext — not viable for real-time DeFi.
This trilemma has been the core technical reason privacy DeFi hasn't worked.
ZK proofs: partial solution
Zero-knowledge proofs solve the problem partially. Tornado Cash proved you could hide transaction origin. Aztec proved you could hide amounts.
But ZK has a ceiling: it proves statements about data without revealing the data, but the data must still be revealed at some point to execute the computation. ZK hides the "who" but struggles to hide the "what" continuously across multiple operations.
FHE: the clean solution
FHE doesn't hide who did something. It hides the data throughout computation. You can:
Run a DeFi protocol on encrypted inputs
The protocol computes the correct output
The output reveals only what the user wants revealed
This is fundamentally different from ZK. FHE enables computation privacy, not just proof privacy.
The performance breakthrough
The reason FHE couldn't solve the trilemma until now was speed. The 2009 Gentry construction was correct but impractical.
TFHE (Torus FHE), developed by researchers at ENS Lyon, changed the picture. Combined with hardware acceleration (AVX-512, GPU acceleration) and protocol-specific optimizations, FHE operations that took minutes now take milliseconds.
Aura benchmarks:
Single encrypted swap: <100ms end-to-end
Throughput: compatible with Solana's TPS targets
Ciphertext expansion: manageable for on-chain storage
The composability question
This is FHE's remaining hard problem: two encrypted values computed by different parties can't easily interact without a coordination mechanism.
Aura's approach is pragmatic: build the encrypted execution layer first, then build the composability bridges. Shield Swap is the proof of concept. The SDK will enable developers to build composable encrypted applications on Aura's runtime.
The full composability picture is a research problem being solved in parallel by the broader FHE community. We expect workable solutions within 18-24 months.
Where we are today
The trilemma is cracking. FHE now delivers:
Privacy: Full computation privacy, mathematically guaranteed
Performance: Fast enough for real DeFi applications
Composability: Limited but growing, with clear research roadmap
Shield.afhe.io is the first DeFi application where all three properties are simultaneously present — even if composability is today limited to the shield ecosystem.
The trilemma isn't solved. But for the first time, you can see the solution clearly.
Live demo: shield.afhe.io | Whitepaper: docs.afhe.io/whitepaper | @AfheLabs