Chainfeeds Introduction:
TEE, MPC, FHE, and ZKP are not competitors. People often try to compare these technologies and determine which is superior. However, in reality, these technologies are not mutually exclusive and can actually work together synergistically, complementing each other's strengths.
Article Source:
https://x.com/paramonoww/status/1930621357479395724
Article Author:
Pavel Paramonov
Perspective:
Pavel Paramonov: MPC protocols are typically divided into three stages: users secret-share their private inputs, send encrypted data to computing nodes, and ensure security through a non-collusion or full-threshold model (where all nodes must collude); then nodes compute these secret shares; finally, nodes return output shares, which users combine to restore the result. MPC is most suitable for networks with good node connectivity, but due to frequent data exchanges between nodes, communication costs are extremely high. In standard MPC protocols, each node needs to communicate with all other nodes, such as during multiplication gate operations, resulting in a communication complexity of O(n²). This means that with 10 nodes, each computing 1 KB, the total data exchange would be around 100 GB; with 100 nodes, it could reach 10 TB. This limits the practical application of MPC to typically 2-10 nodes. In comparison, FHE requires less data transmission but consumes extremely high computational resources. FHE solves the long-standing challenge of "performing secure computations without decryption" - users can upload encrypted data to a server, which performs calculations on the ciphertext, with the output remaining encrypted for user decryption. While preventing data leakage, its computational overhead is massive, with database queries potentially extending from milliseconds to 2-10 seconds, and AI inference from milliseconds to seconds or minutes. In zero-knowledge proofs (ZKP), common uses like zk-rollups typically generate boolean "true/false" proofs, featuring succinctness and quick verification, suitable for on-chain applications. However, zk-rollups do not utilize ZKP's privacy features, as the proof-generating entity can fully access user inputs while running the zk circuit, thus exposing privacy. To address this, zkVM proof processes can be run within a TEE. TEE provides a secure isolated environment, preventing host device access to runtime data while outputting a trustably verified proof. For example, Phala Network uses TEE-enabled GPUs to run SP1 zkVM, with overhead not exceeding 20% even for complex tasks like zkEVM. This combination effectively privatizes zk circuits while maintaining succinctness and verifiability. Additionally, TEE can generate reliable attestation proofs to demonstrate correct circuit execution in a secure environment, enhancing user trust in off-chain proof processes. While TEE is fast and efficient, its reliance on hardware keys challenges data portability and censorship resistance. Combining TEE with MPC can mitigate these issues. For instance, Fairblock uses MPC technology to build confidential computing applications, preventing information leakage and centralization risks. In its architecture, Phala's TEE environment generates private keys, then threshold-encrypts and fragments these keys across Fairblock's MPC node network. Simultaneously, smart contracts monitor TEE operations, requiring periodic encrypted key submissions as a state validation mechanism. If TEE fails, the smart contract triggers the MPC module to privately reconstruct and decrypt the key, ensuring continuous data confidentiality. This architecture achieves dual security: TEE ensures key generation and usage in an isolated environment, while MPC guarantees data recovery and confidentiality during single-point failures. Throughout the process, keys remain encrypted, with no single node able to access the complete key, effectively mitigating centralization risks and enhancing system robustness.
Content Source
