Dashboard

Calibration

Definition. Calibration is the principle under everything Noxtica returns. Noxtica gives you a calibrated risk read — not a verdict. A risk level, a confidence measure, and the reasons behind them. Your team makes the final call; Noxtica gives them better inputs.

A read is something you can reason about: it carries its evidence and its uncertainty. A verdict hides both. Noxtica is built around reads because a blocked real customer never comes back, and you can only defend a decision you can read.

How Noxtica does it

  • A read, not a flag. Every result is a risk level plus a confidence measure plus the reasons — never a single “this is a bot” answer.
  • Transparent statistics. The anomaly checks behind the reads are explainable. Traffic-volume anomaly detection is a trailing-window z-score — it flags hours more than three standard deviations from the rolling mean, conservatively. Population analysis is fleet detection plus hourly rollup histograms and signature matching. These are clear, inspectable statistics — a clear read, not magic — not a trained or learning model.
  • Operator-tuned. You set the thresholds. The defaults are conservative on purpose, and the tuning lives at the threshold level so the calibration stays legible.

The whole approach is “calibration over verdicts” — the first of Noxtica’s engineering principles.

Where we’re headed

This section is forward-looking — a deliberate roadmap note, not a description of today.

Today operators tune policies and thresholds with full control. Self-calibration and feedback loops are being explored for future releases.

To be precise about the line between shipped and planned: today the scoring engine is static and operator-tuned. It does not learn or calibrate itself. Self-calibration — an engine that adjusts its own thresholds from outcomes — is something we’re exploring for the future, not a capability you have now. When we describe Noxtica’s reads in the present tense, we mean the static, operator-tuned engine you can configure today.

Where it fits

Calibration is the principle beneath the whole Agentic Security & Intelligence Platform — it’s why Know Your Device returns a score-with-reasons rather than a verdict, and why the risk-action engine acts only on policies you write.

Read deeper