Effective prediction of Distributed Denial of Service (DDoS) attacks faces several challenges. By preprocessing network traffic for signs of attack preparation, high-dimensional feature sets introduce significant noise due to strong correlations. Selecting relevant features removes noise and improves attack prediction performance. This article advances prediction using Ordinal Pattern Transformation, an approach capable of revealing subtle patterns in network traffic by introducing feature selection as part of the method. One of the outcomes obtained anticipated an attack with 49 minutes and 28 seconds of lead time, reducing the feature set by 94.73%.
If you have any questions or you want to report anything, feel free to reach me at: [email protected] or by my LinkedIn.