Block chain federal learning method and system based on SGX

The invention discloses an SGX-based block chain federal learning method and system, and the method comprises the steps: carrying out the remote authentication between each client and an SGX, and building a safe channel; the client carries out training of a local model and stores parameters of the local model; the client generates a training result and issues the training result to the block chain network; the block chain system uses a consensus mechanism to determine an aggregation server executing aggregation in the current round, the aggregation server asynchronously obtains client model gradients of all clients from the block chain network and sends the client model gradients to an SGX for verification, and after verification succeeds, the SGX performs model aggregation; after completing model aggregation, the SGX generates an aggregation result and issues the aggregation result to the block chain; and other block chain nodes verify the integrity of the aggregation result, add the block to the local block