Epilepsy, characterized by abnormal neuronal discharges, is a widespread neurological disorder. Traditional detection methods primarily focus on extracting time-frequency features from raw electroencephalogram (EEG) signals, often overlooking the complex interactions between brain regions. To address this limitation, we propose an automated approach for epileptic seizure detection. We first construct a fused brain network (TFPP) by integrating the Pearson Correlation Coefficient (PCC) and Phase Lag Index (PLV) across both temporal and frequency domains, enhancing EEG signals by capturing synchronized activities across different brain regions. Next, we introduce the CNN_sk_BiLSTM model, which leverages Convolutional Neural Networks (CNN) and Bidirectional Long Short-Term Memory (BiLSTM) networks to sequentially extract spatial and temporal features. The SKAttention mechanism further refines feature extraction across multiple scales. Using the CHB-MIT scalp EEG dataset, the proposed method achieved an average sensitivity of 90.86%, specificity of 98.79%, and accuracy of 97.88%. These results confirm that fused brain functional networks effectively integrate the intra-channel characteristics of EEG signals, demonstrating the superior performance of the proposed approach in epileptic seizure detection.