1. Feature Representation Models for Cyber Attack Event Extraction
- Author
-
Xiaoxin Lin, Likun Qiu, and Xin Ying Qiu
- Subjects
Context model ,business.industry ,Event (computing) ,Computer science ,Supervised learning ,Feature extraction ,020206 networking & telecommunications ,02 engineering and technology ,010502 geochemistry & geophysics ,Machine learning ,computer.software_genre ,01 natural sciences ,Feature model ,Data modeling ,Feature (computer vision) ,0202 electrical engineering, electronic engineering, information engineering ,Cyber-attack ,Artificial intelligence ,business ,Hidden Markov model ,computer ,0105 earth and related environmental sciences - Abstract
We design and compare multiple feature representation models for classifying cyber attack events and their arguments. Experiment results show that combining lexical, contextual, and semantic features of a sentence performs well for identifying cyber attack event arguments with supervised learning methods and pre-annotated training and test data. However, with implementable simulation experiments with non-annotated test candidates, trigger-matching method works best for event type detection, while word-embedding feature model trained with large corpus performs much better than other models. The comparisons shed lights for future improvement on cyber attack news detection.
- Published
- 2016