Kitsune
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 1
- VX-One paper proposes typed, calibrated decision heads for network perimeter triage
A single-author preprint, VX-One, was submitted to arXiv on 21 October 2026 by Juan S. Santillana, describing a small multi-head network of 100K–10M parameters whose outputs are a calibrated gate, a choice over a closed candidate set that always includes an explicit unknown option, and an ordinal score, at roughly 0.7–1.7 ms median latency on CPU. In the author's own experiments on UNSW-NB15, CIC-IDS2017 and EMBER2024 against calibrated XGBoost, Random-Forest, MLP and Kitsune, VX-One reportedly wins detection (AUROC/FPR@95%TPR) on UNSW-NB15 and under temporal drift on CIC-IDS2017 (0.711 vs 0.637), while XGBoost wins multi-class attribution on all three benchmarks. The paper also reports that calibration collapses under temporal drift (ECE 0.04 to 0.40) for every method, that abstention thresholds transfer poorly across benchmarks, and that on payload-less reconstructions signature engines resolve only about 0.1% of flows, leaving a gray zone the second cascade phase recovers to combined recall of at least 0.978.
Hugging Face · Papers · 🔥 0
Experience and discussion from the community
Share my Kitsune experienceAsk about Kitsune
Nobody has shared their experience with Kitsune yet.