KDD Cup '99 - SMTP

The original KDD Cup '99 dataset from UCI machine learning repository contains 41 attributes (34 continuous and 7 categorical). This variant only includes observations with "smtp" as their service feature and strips the protocol type and flag categorical features. All remaining features are included as-is.

KDD Cup '99 (HTTP) is available on Aftershock and normal observations are available in the included training dataset consisting of 18 dimensions per observation. During evaluation, the main program of your submission is expected to access /ingress/kddcup99_smtp/testing.csv which has the same form as the development dataset. You are then expected to produce a single-column CSV file of anomaly confidence values (in [0, 1]) at /egress/kddcup99_smtp/predictions.csv. Your predictions must be sequentially aligned with the rows of the testing dataset.

Scores

Method Authors AUC Runtime (ms)
Bionic (Pre-Release) K. Demetriou, I. Becker, S. Hailes 0.999 1,864.026
Baselines - Isolation Forest K. Demetriou, I. Becker, S. Hailes 0.999 3,294.884
Baselines - One Class SVM K. Demetriou, I. Becker, S. Hailes 0.778 35,834.490
Baselines - Robust Covariance K. Demetriou, I. Becker, S. Hailes 0.947 47.054
Baselines - Local Outlier Factor K. Demetriou, I. Becker, S. Hailes 0.503 6,309.263