A Robust Cybersecurity Topic Classification Tool

In this research, we use user defined labels from three internet text sources (Reddit, StackExchange, Arxiv) to train 21 different machine learning models for the topic classification task of detecting cybersecurity discussions in natural English text. We analyze false positive and negative rates each model’s cross validation experiments. Then present a Cybersecurity Topic Classification (CTC) tool, which takes majority vote trained as decision mechanism related also show that CTC tool provides lower on average than any individual models. is scalable hundreds thousands documents with wall clock time order hours.