Abstract: To address imbalanced data challenges in intrusion detection, we propose SA-LCA, integrating an improved stacked autoencoder with LSTM-CNN-Attention. Preprocessing uses one-hot encoding and ...