Fair Preprocessing Re-Scores Three Lateral Movement Detectors

Three well-known lateral-movement detectors scored differently once a paper reran them under matched preprocessing rules, per arXiv 2607.29390. That's a problem for anyone who picked a detection tool off a benchmark table built from these two datasets.

The paper targets methodology, not the detectors themselves. On the two datasets most lateral-movement research relies on, preprocessing and event-labeling methodology varies substantially paper to paper. That choice quietly decides how fair the resulting evaluation is. Standardize it and the same detectors don't land where their original papers put them.

The paper doesn't say which direction each detector moved, or name the three detectors specifically. That's a gap if you're trying to map this straight to a purchase decision, but it doesn't undercut the methodology finding.

The fix isn't a better detector, it's a standard labeling spec for these two datasets that every paper has to use. Watch whether that spec shows up. Until it does, every leaderboard built on these datasets stays provisional.