@benj23
the main problems the team @zama might face:
the computational load of FHE is quite a serious obstacle since it’s catastrophically expensive and pretty slow. even with optimizations and powerful nodes, there’s a category of operations that remain heavy. this problem can be solved by certain high performance coprocessors and simple schemes (bootstrapping, noise management)
threshold/distributed decryption - so that no single party can decrypt the shared key, it’s sharded among operators; this adds extra delays and risks of collusion, especially if the node set is small
verification and proofs - it’s necessary to prove the correctness of the coprocessor’s work (otherwise results could be faked), and this is usually done “hybrid style” with ZK approaches or MPC assertions. these issues are discussed in engineering and analytical papers on the topic
0 reply
0 recast
0 reaction