Fig 1.
FLDP intrusion detection framework and model pruning process.
Fig 2.
Illustration of the process of the intrusion detection model.
Firstly, the local private data is grouped, and then these data are formatted into a time series format to meet the training requirements of the time series model. The formatted data is then inputted into a Multilayer Perceptron (MLP) to generate the final detection results.
Fig 3.
Experimental setup.
Table 1.
Federated learning for training model parameters and hyperparameters.
Fig 4.
Performance evaluation of the FFIDS algorithm.
Evaluation of different datasets with various pruning rates and 150 epochs iteration.
Table 2.
Communication cost of different models on different datasets and pruning rates.
Table 3.
Communication and time cost on FL vehicle-level training.
Fig 5.
Comparing the F1 score of different baseline models.
Comparing each global iteration with a fixed pruning rate of 50 with 150 epochs iteration.
Fig 6.
Comparing the F1 score of different baseline models.
Comparing by varying privacy budgets from 6 to 16 with 150 epochs iteration.