Can Adversaries Game
a Static Baseline?

Watch a sophisticated adversary learn the CAD threshold and stay just below it — then see ML-based adaptive windowing close the detection gap in real time.

Events

Detected

Evaded

0.65

Threshold

ML Confidence

CAD ANOMALY SCORE — TIMELINE

Normal

Adversary

Static threshold

Adaptive threshold

READYSelect a mode and press RUN to begin simulation.

SIMULATION CONTROLS

Adversary Learning Rate 0.4

Adversary Aggression 0.6

Decay Window (events) 10

ML MODEL — LIVE PARAMETERS

Pattern Recognition0%

Frequency Anomaly0%

Cluster Deviation0%

Threshold Pressure0%

● ADVERSARY STATUS

STRATEGYProbing

LEARNED THRESHOLDUnknown

CURRENT SCORE—

EVASION STATUSMONITORING

EVENTS UNTIL DETECT—

Mode 01 — Static Threshold: The adversary probes with high anomaly scores, observes non-detection, and learns the static threshold. Once learned, it stays just below — evading indefinitely. This is the fundamental weakness of fixed baselines.

Can Adversaries Gamea Static Baseline?

Can Adversaries Game
a Static Baseline?