VX-One：面向网络边界分流的类型化校准决策头研究
VX-One: Typed, Calibrated Decision Heads for Real-Time Perimeter Triage - An empirical study of unknown-attack honesty, cascade gray zones, and calibration under drift
一个把网络攻击分流拆成三个校准小头的小模型，CPU 毫秒级跑完，还老老实实列出了哪些场景翻车，做安全检测的可以看看。
arXiv 论文提出 VX-One，一个 100K-10M 参数的小型多头网络，输出二值门控、含未知选项的闭集分类和序数分数三种类型化决策原语，CPU 上 p50 延迟约 0.7-1.7 ms。在 UNSW-NB15、CIC-IDS2017、EMBER2024 三个基准上与校准后的 XGBoost、Kitsune 等对比，VX-One 在 UNSW-NB15 和 CIC-IDS2017 时间漂移场景下检测占优（AUROC 0.711 vs 0.637），但 XGBoost 在三个基准的多类归因上均胜出。对 Suricata 52674 条 Emerging-Threats 规则加 VX-One 的两阶段级联做五天 CIC-IDS2017 测试，无负载重建流量中签名引擎只能解析约 0.1%，级联组合召回不低于 0.978。论文也实测了局限：留一家攻击族交叉验证在 9/9 个 UNSW 折上可检出未知族，但在 CIC recon 族上低于随机，时间漂移下所有方法的 ECE 从 0.04 恶化到 0.40。
VX-One: Typed, Calibrated Decision Heads for Real-Time Perimeter Triage - An empirical study of unknown-attack honesty, cascade gray zones, and calibration under drift
Perimeter triage faces a latency/semantics trade-off: signature engines are fast and format-safe but blind to unseen attacks, while generative models are semantic but slow and produce unparseable output. In late 2026 a class of commercial System One decision models (TypeSafe Jev; Microsoft-Decision-1) began selling typed decision primitives - binary gates, closed-set choices, ordinal scores - as product APIs, with no academic treatment. We present VX-One: a small multi-head network (100K-10M parameters) whose outputs are these primitives - a calibrated gate, a choice over a closed candidate set that always includes an explicit unknown option, and an ordinal score, trained multi-task, temperature-calibrated per head (~0.7-1.7 ms p50 on CPU). We evaluate it on UNSW-NB15, CIC-IDS2017, EMBER2024 against calibrated XGBoost, Random-Forest, MLP, Kitsune, and measure a two-phase cascade (Suricata with 52,674 Emerging-Threats rules -> VX-One) over five days of CIC-IDS2017. Findings: (i) VX-One wins detection (AUROC/FPR@95%TPR) on UNSW-NB15 and under temporal drift on CIC-IDS2017 (0.711 vs 0.637), but XGBoost wins multi-class attribution on all three benchmarks; (ii) leave-one-family-out gate detects never-seen families on 9/9 UNSW folds yet drops below chance on CIC recon family, and abstention threshold transfers poorly across benchmarks; (iii) on payload-less reconstructions signature engines resolve ~0.1% of flows - gray zone ~99.9% of traffic, recovered by Phase 2 (combined recall >=0.978); (iv) calibration collapses under temporal drift (ECE 0.04->0.40) for every method while ranking partially survives. We provide a measured account of where typed calibrated decisions help perimeter triage, and where they fail.
