FLDID: Federated Learning Enabled Deep Intrusion Detection in Smart Manufacturing Industries

. 2022 Nov 19;22(22):8974. doi: 10.3390/s22228974

Algorithm 1 Secured FL procedure

Input: Set of Edge devices

E_{n}

and their associated local data

D_{n} | n \in N

, No. of communication rounds R

Output: Global DID model

1: Cloud server initialize model parameter

W_{0}

and

η, B, ζ, ρ, τ

2: Each

E_{n}

informs the data size

| D_{n} |

to server;

3: Server computes contribution ratio

α_{n} = | D_{n} | / (| D_{1} | + | D_{2} | + \dots + | D_{N} |)

;

4: For each

E_{n}

KMC generates key pair

(P u b K, P r i K)

using

K e y G e n (n)

and establishes secure channel between each edge device and server;

5: For r = 1 to R;

(i)

E_{n}

computes

r^{t h}

round model gradients

W_{n}^{r}

using Algorithm 2;

(ii) Encrypt model gradients

W_{n}^{r}

E (W_{n}^{r}) = E n c r y p t_g r a d (W_{n}^{r}, P u b K)

;

(iii) Upload

E (W_{n}^{r})

to cloud server;

End

6: Cloud server aggregates

E (W_{n}^{r})

C = A g g r_g r a d (E (W_{1}^{r}), (W_{2}^{r}), \dots α_{1}, \dots α_{n})

;

7: Server distributes C to all edge devices

E_{n}

8: To obtain new global model

E_{n}

performs decryption as

\tilde{W^{r}} = D e c r y p t_g r a d (C, P r i K)

E_{n}

updates model gradients using

\tilde{W^{r}}

;

10:

r \leftarrow r + 1

;