In the realm of cybersecurity knowledge graph construction, this study introduces a cutting-edge named entity recognition framework that innovatively combines the BERT-FFN-BiLSTM-CRF architecture. Anchored by the BERT pre-trained model, this method excels at drawing out intricate semantic vectors from textual inputs, offering a robust representation of language. It then employs a Feed-Forward Neural Network (FFN) for fine-tuning these vectors, sharpening the detection of distinct entity traits. Additionally, it leverages the Bidirectional Long Short-Term Memory (BiLSTM) network to thoroughly explore contextual interrelations within the text. Finally, a Conditional Random Field (CRF) layer is applied to meticulously execute entity tagging and sequence decoding tasks. Verified against a specialized dataset for named entity recognition in cybersecurity, the model's experimental results indicate substantial enhancements in crucial metrics—namely accuracy, recall, and the F1 score—marking a pronounced superiority over comparative models.