This website requires JavaScript.

A Robust Cybersecurity Topic Classification Tool

Elijah PelofskeLorie M. LiebrockVincent Urias
摘要
In this research, we use user defined labels from three internet text sources(Reddit, Stackexchange, Arxiv) to train 21 different machine learning modelsfor the topic classification task of detecting cybersecurity discussions innatural text. We analyze the false positive and false negative rates of each ofthe 21 model's in a cross validation experiment. Then we present aCybersecurity Topic Classification (CTC) tool, which takes the majority vote ofthe 21 trained machine learning models as the decision mechanism for detectingcybersecurity related text. We also show that the majority vote mechanism ofthe CTC tool provides lower false negative and false positive rates on averagethan any of the 21 individual models. We show that the CTC tool is scalable tothe hundreds of thousands of documents with a wall clock time on the order ofhours.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答