A Cybersecurity Entity Recognition Method for Enhancing Situation Awareness in Power Systems
Chenwei Yang, Youfeng Niu, Hao Huang, Shilong Zhang, Xiaozhi Deng, Yunfan Yang, Qinqin Wu, Yang Liu · 2024
The intrusion detection system (IDS) within the power system detects network anomalies and activates alarms by matching rules. To address the evolving landscape of network attack methods, continuous updates to the ruleset are necessary. The integration of threat intelligence knowledge into the ruleset is crucial for enhancing the system’s security awareness and response capabilities. The prevailing approach involves leveraging NLP techniques for intelligence extraction, in which NER is a key step. Due to the complexity and variability of entity names, achieving high-precision NER is challenging. In this study, a novel method that combines RoBERTa, Bi-GRU, and CRF is proposed. By utilizing RoBERTa’s comprehensive semantic word vectors and Bi-GRU embedded with attention mechanisms, the method enhances feature extraction capabilities. This approach aims to achieve high-precision recognition of security-related entities, thereby enhancing the system’s ability to extract valuable insights from threat intelligence.