FL-DSFA: Securing RPL-Based IoT Networks against Selective Forwarding Attacks Using Federated Learning

. 2024 Sep 8;24(17):5834. doi: 10.3390/s24175834

Algorithm 1 Federated learning with Selective Forwarding Attack Detection.

1:
Input: Local datasets ${D_{1}, D_{2}, \dots, D_{n}}$ for each CPS node
2:
Output: Global Model $M_{global}$
3:
Initialization:
4:
Set communication rounds T, local models M, learning rate $η$ , aggregation method $ϕ$
5:
Initialize global model $M_{global}^{(0)}$
6:
Fog Layer Processing:
7:
for $t = 1$ to T do
8:
Receive local models from CPS nodes: ${M_{1}^{(t)}, M_{2}^{(t)}, \dots, M_{n}^{(t)}}$
9:
Aggregate models: $M_{global}^{(t)} = ϕ ({M_{1}^{(t)}, M_{2}^{(t)}, \dots, M_{n}^{(t)}})$
10:
Broadcast global model $M_{global}^{(t)}$ to all CPS nodes
11:
end for
12:
CPS Layer for Local Model Training:
13:
for $i = 1$ to n do
14:
Initialize local model $M_{i}^{(0)}$
15:
for $t = 1$ to T do
16:
Receive global model $M_{global}^{(t)}$
17:
Train local model $M_{i}^{(t)} = Train (M_{i}^{(t - 1)}, D_{i})$ using KNN, LR, SVM, NB
18:
Transmit local model $M_{i}^{(t)}$ to Fog layer
19:
end for
20:
end for
21:
Final Aggregation:
22:
Finalize the global model $M_{global}$