ChronoSentinel: Incremental temporal embedding for Security Knowledge Graph using Dynamic Reachability Centrality and Efficient language model

Chinmaya Mishra, Himangshu Sarma, Saravanan M.

Abstract: Highlights•Cybersecurity: Applying knowledge graphs for threat detection and prediction.•Knowledge Graphs (KGs): Using Security Knowledge Graphs (SKGs) to model and analyze relationships among malware, vulnerabilities, and attack patterns.•Temporal Knowledge Graphs (T-KGs): Integrating time-sensitive data into SKGs to capture the evolution of cyber threats.•Graph-Based Learning and Ranking: Utilizing Dynamic Recahability Centrality to prioritize core nodes based on their temporal relevance.•Machine Learning (ML) and Language Model (LM): Employing Efficient Language Models (ELMs) such as BART, FLAN-T5, and DeepSeek to enhance graph quality and contextual understanding.

External IDs:doi:10.1016/j.inffus.2025.103662