BFL-SDWANTrust: Blockchain Federated-Learning-Enabled Trust Framework for Secure East–West Communication in Multi-Controller SD-WANs

View full-text article in PMC

. 2025 Aug 21;25(16):5188. doi: 10.3390/s25165188

Algorithm 4 Cloud aggregation layer algorithm for global trust model aggregation

Require: Set of controllers $C = {C_{1}, \dots, C_{N}}$ , encrypted local models ${C_{i}}$ , encryption

keys ${K_{i}}$ , decryption keys ${K_{i}^{- 1}}$ , threshold $θ$ , telemetry volumes ${v_{i}}$
Ensure: Encrypted global trust models ${C_{g l o b a l}^{i}}$
1: Compute total volume $V \leftarrow \sum_{i = 1}^{N} v_{i}$
2: Initialize verified models $V \leftarrow \emptyset$ , volumes $W \leftarrow \emptyset$
3: for $i = 1$ to N do
4: Receive encrypted model $C_{i}$
5: Decrypt ${\hat{T}}_{i} \leftarrow Decrypt (C_{i}, K_{i}^{- 1})$
6: if Signature verification of ${\hat{T}}_{i}$ succeeds then
7: Add ${\hat{T}}_{i}$ to $V$
8: Add $v_{i}$ to $W$
9: else
10: Discard ${\hat{T}}_{i}$ and flag $C_{i}$
11: end if
12: end for
13: Compute weighted average gradient:
$\bar{g} \leftarrow \frac{1}{\sum w_{j}} \sum_{j} w_{j} \cdot {\hat{T}}_{j}, {\hat{T}}_{j} \in V, w_{j} \in W$
14: for each ${\hat{T}}_{i} \in V$ do
15: Compute deviation $d_{i} \leftarrow ∥ {\hat{T}}_{i} - \bar{g} ∥$
16: if $d_{i} > θ$ then
17: Flag ${\hat{T}}_{i}$ as malicious
18: Remove ${\hat{T}}_{i}$ and corresponding $w_{i}$ from $V, W$
19: end if
20: end for
21: Recompute total volume $V \leftarrow \sum w_{j}$
22: Aggregate global model:
$T_{g l o b a l}^{t + 1} \leftarrow \sum_{j} \frac{w_{j}}{V} \cdot {\hat{T}}_{j}$
23: for $i = 1$ to N do
24: Encrypt global model for controller i:
$C_{g l o b a l}^{i} \leftarrow Encrypt (T_{g l o b a l}^{t + 1}, K_{i})$
25: Transmit $C_{g l o b a l}^{i}$ to $C_{i}$
26: end for
27: return ${C_{g l o b a l}^{i}}$