Predicting Cyber Events by Leveraging Hacker Sentiment

摘要: Recent high-profile cyber attacks exemplify why organizations need better defenses. Cyber threats are hard to accurately predict because attackers usually try mask their traces. However, they often discuss exploits and techniques on hacking forums. The community behavior of the hackers may provide insights into groups' collective malicious activity. We propose a novel approach events using sentiment analysis. test our attack data from 2 major business organizations. consider 3 types events: software installation, destination visits, emails that surpassed target organizations' construct predictive signals by applying analysis hacker forum posts understand behavior. analyze over 400K generated between January 2016 2018 100 forums both surface Dark Web. find some have significantly more power than others. Sentiment-based models leverage specific can outperform state-of-the-art deep learning time-series forecasting weeks ahead events.