Output details

Exploring discrepancies in findings obtained with the KDD Cup '99 data set

This paper explores important issues with a breadth of literature, describing and then seeking to explain the many contradictory results, all of which are based on analysis of the same important dataset. The perspective is intrusion detection, and the paper ultimately seeks to improve techniques to detect intrusions, and alleviate costly attacks on networks, such as Denial of Service. The extensive and rigorous treatment not only brings lessons learned which would apply to other areas of imbalanced data, such as fraud, medicine, or customer churn, but also has implications for documentation of experimental methods that proliferate the machine learning literature.