COMPARISON OF IOB1 AND IOE TAGGING SCHEMES IN NAMED ENTITY RECOGNITION FOR ADVANCED PERSISTENT THREAT ENTITY EXTRACTION USING BILSTM

This poster presents a study on the application of Named Entity Recognition (NER) using a BiLSTM-based model to extract important entities from Advanced Persistent Threat (APT) reports in the Cyber Threat Intelligence (CTI) domain, by comparing two tagging schemes, namely IOB1 and IOE. The study utilizes the CyberNER dataset, which consists of 6,311 sentences and 204,815 tokens. The research methodology includes data collection, tagging scheme transformation, data preprocessing, data splitting, model development and training, as well as evaluation using performance metrics. The results demonstrate excellent performance, where IOB1 achieves an F1-score of 94.87% and IOE achieves 94.29%. Overall, this approach enhances the accuracy of automated entity extraction in CTI.