RFID Scheme for IoT Devices Based on LSTM-CNN

Algorithm 1

RFID algorithm based on LSTM-CNN.

Input: RF data set

Offline training stage:

1. Slice to obtain RF signal subsequences set , =[2, 128]

2. Normalize to get for the input of the LSTM-CNN model

3. Add label to the data set,

4. Training process:

(1) Input and labels into the LSTM units for eigenvectors set with temporal information

(2) Input into the CNN network to obtain the classification result on the training set

(3) Adjust the network model parameters according to loss function to obtain a fully trained LSTM-CNN model

Online identification stage:

for new RF data do:

Get the predictive probability set with trained authentication model,

ifthen

, the label type corresponding to is the classification result

else

Judging that the identity of the unknown node is illegal

Output: identity type of the unknown node.