Table 7 Ablation study on key components of the proposed framework.

From: An efficient data driven framework for intrusion detection in wireless sensor networks using deep learning

Model variant

Accuracy

Precision

Recall

F1-score

Full Model

97.4%

96.3%

95.9%

96.1%

Without Adversarial Training

94.2%

92.7%

91.1%

91.8%

Without SMOTE

93.6%

92.1%

89.3%

90.6%

Without RNN Module

92.9%

91.5%

89.7%

90.6%

CNN Only (No RNN, SMOTE, Adv)

90.1%

89.2%

86.8%

88.0%