Mask-Based Window-Level Insider Threat Detection for Campaign Discovery
Jericho Cain*, Hayden Beadles · 2026
Read Mask-Based Window-Level Insider Threat Detection for Campaign Discovery by Jericho Cain*, Hayden Beadles (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-11019.txt.
Description
Insider attacks hide in long stretches of quiet audit logs. Hayden Beadles and Jericho Cain split each time window into two parts. A mask says whether activity happened, and values say how strong it was. A two-channel autoencoder rebuilds both, with value loss counted only where activity exists.
Tests on the CERT r5.2 set show short attacks surface mainly through presence, while longer ones add a magnitude signal. Noise weakens magnitude and pushes finding back to presence. At campaign scale, signal piles into a few odd windows, so simple pooling of extreme scores recovers long attacks without sequence models. The lesson is that matching the model to sparse time structure matters more than adding hardness.
Catalog record
| CREATOR | Jericho Cain*, Hayden Beadles |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 43 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-11019