federated learning and ECC for attack detection in IIoT

I am writing a program to improve privacy and security in IIoT. In this program, we have n clients, which are IIoT devices. The main goal of this program is to detect anomalies and attacks in IIoT. For this purpose, we used the Swat dataset. Now we are using federated learning. In federated learning, each client trains the local model with their data and then shares it with central server to detect anomalies. The central server for detecting anomalies uses an auto-encoder decoder model. To improve privacy, we use Elliptic-curve cryptography. When each client wants to share their local model with the central server, first encrypt it with the public key of the central server to make sure no one else can access the model. When the central server receives the models from clients, it first encrypts them with its private key, then aggregates and make a global model. Then when a central server wants to send the global model to clients for the update, apply its digital signature. Clients decrypt the digital signature when they receive the model to ensure this is from the central server. This process continues until the model can detect anomalies in IIoT with high performance. We evaluate our model using Swat, gas pipeline, and unsw nb15 datasets. Required skill: Advanced machine learning and cryptography.