Cyber Attacks Against Enterprise Networks: Characterization, Modeling and Forecasting

Cyber attacks are a major and routine threat to the modern society. This highlights the importance of forecasting (i.e., predicting) cyber attacks, just like weather forecasting in the real world. In this paper, we present a study on characterizing, modeling and forecasting the number of cyber attacks at an aggregate level by leveraging a high-quality, publicly-available dataset of cyber attacks against enterprise networks; the dataset is of high quality because more than 99% of the attacks were examined and confirmed by human analysts. We find that the attacks exhibit high volatilities and burstiness. These properties guide us to design statistical models to accurately forecast cyber attacksand draw useful insights.