Advancing Cybersecurity through the Development of a Semantic Knowledge Base: Novel Methods and Schemes for Multi-source Data Integration

Abstract

The paper presents a novel method for constructing a multi-domain, multi-layer cybersecurity semantic knowledge base. Utilizing diverse cybersecurity data sources, the framework seamlessly integrates entity extraction, event extraction, and semantic relation extraction through sophisticated semantic parsing techniques, resulting in a highly accurate and comprehensive knowledge base. By employing Bi-directional Long Short-Term Memory (Bi-LSTM) models with attention mechanisms for precise entity extraction, a hierarchical strategy network for detailed event extraction, and pattern rule matching for elaborate semantic relation extraction, the proposed method demonstrates exceptional efficacy. Experimental results show entity extraction with 95.4% accuracy and 92.5% recall, event extraction with 93.3% accuracy and 90.8% recall, and semantic relation extraction with 90.7% accuracy and 88.6% recall. The constructed knowledge base achieves an average query response time of 0.5 s and a query accuracy of 92%. This innovative approach enhances the processing and understanding of complex cybersecurity data, providing reliable and precise semantic knowledge query and reasoning, crucial for dynamic threat response.