Privacy-Preserving Detection of Tampered Radio-Frequency Transmissions Utilizing Federated Learning in LoRa Networks

. 2024 Nov 17;24(22):7336. doi: 10.3390/s24227336

Algorithm 1 The use of pseudocode in Multi-Model Federated Learning (FL) for the detection of anomalies.

1:
$D \leftarrow L o a d D a t a ()$ ▹ Load the dataset from designated directories.
2:
if any $d_{k} \in D$ is missing or invalid, apply imputation
3:
Convert D into a feature matrix: $X = [x_{1}, x_{2}, \dots, x_{n}]$
4:
Perform normalization on the data. X: $X = n o r m a l i z e (X)$
5:
Split X into training set $X_{t r a i n}$ and validation set $X_{v a l}$
6:
$c l i e n t s \leftarrow 5$ ▹ Number of clients
7:
$m o d e l s \leftarrow [O C S V M, C A E, I s o l a t i o n F o r e s t, L O F, K M e a n s]$ ▹ Array of candidate models
8:
$c l i e n t_d a t a s e t s \leftarrow s p l i t_d a t a_f o r_c l i e n t s (X_{t r a i n}, c l i e n t s)$
9:
for each client i do
10:
for each model $m \in m o d e l s$ do
11:
Train model m on $c l i e n t_d a t a s e t s [i]$
12:
Store the trained model $m_{i}$
13:
end for
14:
end for
15:
$a g g r e g a t e d_m o d e l s \leftarrow []$ ▹ Array to store aggregated models
16:
for each model $m \in m o d e l s$ do
17:
Aggregate all client models for m into a global model $m_{g l o b a l}$
18:
Append $m_{g l o b a l}$ to $a g g r e g a t e d_m o d e l s$
19:
end for
20:
for each $m_{g l o b a l} \in a g g r e g a t e d_m o d e l s$ do
21:
Evaluate $m_{g l o b a l}$ on $X_{v a l}$
22:
Calculate and present the evaluation metrics: accuracy, precision, recall, and F1-score.)
23:
end for