Scanner/runs in this tab
JavaScript engine · checks L1–L6
Scan your own split
Exact and near copies, label conflicts, time leakage, one-column shortcuts and group overlap between a training and a test file.
Your files never leave this page.
The page's Content-Security-Policy sets connect-src 'none', so it cannot make network requests, and the scan runs in a worker that inherits that policy. Check it yourself: open your browser's network panel and run a scan.
Sized for splits up to about half a million rows (measured: 500,000 rows of 20 columns in about 2 minutes and 2.8 GB of memory). For larger datasets, such as KDD Cup 99's 4.9 million rows, use the command-line tool or its Docker image, which run the same checks with NumPy.
1
CSV with a header row · same columns in both
Load the two files
TRAIN SPLIT
Loaded
TEST SPLIT
Loaded
Or try an example:
2
Load both files to continue
Tell it about the columns
Leave out of the features (IDs, the attack name, anything that encodes the label)
—
Ctrl + Enter