Table 7 Performance comparison of FedT-DQN under heterogeneous data distribution scenarios.
From: A federated transformer-enhanced double Q-network for collaborative intrusion detection
Dataset | ACC | AUC | Prec. | Rec. | F1 |
|---|---|---|---|---|---|
Random distribution | |||||
 CICIDS2018-v2 | 0.9945 | 0.9918 | 0.9811 | 0.9781 | 0.9987 |
 BoT-IoT-v2 | 0.9901 | 0.9945 | 0.9841 | 0.9838 | 0.9936 |
 ToN-IoT-v2 | 0.9871 | 0.9941 | 0.9414 | 0.9814 | 0.9710 |
 UNSW-NB15-v2 | 0.9881 | 0.9945 | 0.9488 | 0.9996 | 0.9774 |
Customized distribution | |||||
 CICIDS2018-v2 | 0.9922 | 0.9942 | 0.9679 | 0.9564 | 0.9594 |
 BoT-IoT-v2 | 0.9987 | 0.9936 | 0.9978 | 0.9951 | 0.9993 |
 ToN-IoT-v2 | 0.9984 | 0.9936 | 0.9384 | 0.9994 | 0.9622 |
 UNSW-NB15-v2 | 0.9957 | 0.9992 | 0.9551 | 0.9921 | 0.9522 |
