Instrument 1.0 Domain SOC human factors Signal-detection · workload · latency Participants anonymous PII none
Human factors in security operations

Measuring how analysts separate signal from noise.

foyl Research is a calibrated testbed for attention, workload, and decision-making in the SOC. It runs realistic analyst tasks as controlled experiments, manipulating noise, time pressure, and framing while logging millisecond-level behavior, so human-factors questions can be measured, not asserted.

or read the method ↓
Fig 1 · Signal-detection model d′ = 1.62 NOISE SIGNAL criterion c d′
01

The signal chain

Every action passes through a logging pipeline, parameterised by an experimental condition engine.
STAGE01

Simulation engine

Realistic SOC tasks, alert triage and phishing detection, set in a fictional foyl SecIntel world.

out → stimulus
STAGE02

Telemetry pipeline

Timestamped action stream, time-to-first-diagnostic, response time, and friction: hesitation, backtracking, context-switching.

append-only · idempotent
STAGE03

Condition engine

Signal noise, time pressure, and framing, assigned between-subjects and overridable for directed runs.

randomised · 3 factors
02

What the instrument reads

Behavioural measures and validated instruments, per task and per condition.
SymbolMeasureWhat it captures
d′SensitivityDiscriminability of threat from benign, from hits and false alarms (signal-detection theory).
cCriterionDecision bias: how readily the analyst escalates, independent of sensitivity.
t₀Time-to-first-diagnosticLatency from stimulus onset to the first committed decision.
FrictionHesitation, verdict backtracking, and context-switching across the task.
TLXWorkloadNASA-TLX subscales rated after each task.
SESelf-efficacyShort security self-efficacy scale, measured pre and post.
03

Three dials, one participant

Randomised between-subjects. Override any dial with a URL parameter for a directed run.
Signal noise
Alert fatigue · discriminability
LowHigh
Time pressure
Speed–accuracy tradeoff
OffOn
Framing
Self-efficacy · motivation
ControlMastery
Research capability · experimental control

Configurable experimental toggles for manipulating environmental variables (signal-to-noise ratio, time pressure, and feedback mechanics) across control and treatment groups, with per-participant randomised assignment and deterministic, URL-addressable condition cells for reproducible human-subjects runs.

A session takes about 12 minutes and records no personal data.

No account · no email · anonymous participant code · every scenario fictional
Begin a session →

A technical report on the instrument and pilot results is in preparation. Researchers can review the aggregate dashboard and study information.