Network Intrusion Classifier

Live Engine · RandomForest

Classify network connections as normal or attack — scored honestly on held-out NSL-KDD

A real scikit-learn RandomForest is trained in the backend on a labelled sample of NSL-KDD network connections, then classifies a held-out test set it has never seen into five classes: normal, DoS, Probe, R2L and U2R. The model never sees the test labels, so everything below is scored honestly against ground truth.

What this is: a genuine supervised classifier (scikit-learn RandomForest) over the NSL-KDD intrusion-detection benchmark (Canadian Institute for Cybersecurity, UNB). It is trained live on a bundled labelled sample and scored on a held-out test split. NSL-KDD is a dated benchmark (late-1990s attack families), so these numbers do not transfer directly to modern live traffic — and R2L and U2R are famously rare and hard, which the per-class recall shows honestly rather than averaging it away.