Table 2 Implementation parameters for federated learning, block chain, and homomorphic encryption.
Parameter | Component | Description / configuration |
|---|---|---|
Number of FL clients | Federated learning | 5 Jetson Nano edge nodes connected via MQTT over TLS 1.3 |
Data distribution | Federated learning | Heterogeneous (60% normal, 20% mild anomaly, 20% severe anomaly) |
Local epochs / rounds | Federated learning | 3 local epochs per round; 10 global rounds |
Communication protocol | Federated learning | MQTT with TLS encryption and 25% payload compression |
Aggregation method | Federated learning | FedAvg (weighted mean based on sample count) |
Encryption scheme | Homomorphic encryption | Paillier (public key n = 1024 bits, g = n + 1) |
Encryption library | Homomorphic encryption | Python Paillier (PyPaillier v1.5) |
Computation overhead | Homomorphic encryption | ≈ 10.1% inference delay vs. plaintext FL |
Block chain framework | Block chain | Hyperledger Besu Proof-of-Authority (private network) |
Validator nodes | Block chain | 3 validators (hosted on edge gateway servers) |
Block time / throughput | Block chain | 5 s block interval; ≈ 180 tps transaction rate |
Smart contract usage | Block chain | Audit logging and model-update verification |
Hash algorithm | Block chain | SHA-256 for transaction integrity |
