CyTIE: Cyber Threat Intelligence Extraction with Named Entity Recognition

CyTIE: Cyber Threat Intelligence Extraction with Named Entity Recognition

연구 분야: Safety

학회: International Conference on Advancements in Smart Computing and Information Security

In the dynamic intersection of Natural Language Processing and cyber security, Named Entity Recognition plays a pivotal role in comprehending and countering cyber threats. This paper explores Named Entity Recognition techniques within the cyber security context, utilizing a meticulously curated dataset with 12 distinct entity types extracted from security blogs. Our study involves developing and comparative analysis of five Named Entity Recognition models: BiLSTM, BiLSTM-CRF, BERT, BERT-CRF, and BERT-BiLSTM-CRF. Rigorous evaluation reveals that the BERT-BiLSTM-CRF model outperforms others with an F1-Score of 0.9635, excelling at extracting entities from the intricate language used in cyber security texts. Through this paper, we contribute to the ongoing Named Entity Recognition discourse in cyber security, paving the way for advancements in Natural Language Processing techniques and fortifying cyber security measures against evolving digital threats. The implementation and dataset are accessible on our Github page: https://github.com/OPTIMA-CTI/CyberNER.git.