Cyber Weather Forecasting: Forecasting Unknown Internet Worms Using Randomness Analysis

Since early responses are crucial to reduce the damage from unknown Internet attacks, our first consideration while developing a defense mechanism can be on time e ciency and observing (and predicting) the change of network statuses, even at the sacrifice of accuracy. In the recent security field, it is an earnest desire that a new mechanism to predict unknown future Internet attacks needs to be developed. This motivates us to study forecasting toward future Internet atacks, which is referred to as CWF (Cyber Weather Forecasting). In this paper, in order to show that the principle of CWF can be realized in the real-world, we propose a forecasting mechanism called FORE (FOrecasting using REgression analysis) through the real-time analysis of the randomness in the network tra c. FORE responds against unknown worms 1.8 times faster than the early detection mechanism, named ADUR (Anomaly Detection Using Randomness check), that can detect the worm when only one percent of total number of vulnerable hosts are infected. Furthermore, FORE can give us timely information about the process of the change of the current network situation. Evaluation results demonstrate the prediction e ciency of the proposed mechanism, including the ability to predict worm behaviors starting from 0.03 percent infection. To our best knowledge, this is the first study to achieve the prediction of future Internet attacks.