Research questions
- Which explicitly supported scanner configurations produce an observable signature that Flock Block detects?
- How often does the device alert when each supported target is active, inactive, or configured for passive-only collection?
- How often do ordinary non-target devices create false alerts?
- How does the observable radio environment change with the decoy function off and on?
- How do distance, obstruction, orientation, congestion, and backhaul method affect results?
Target matrix and ground truth
Every target record must name manufacturer, model, hardware revision, firmware or software version, radios, collection mode, backhaul, signature-library version, and test date. Ground truth records exact power, scan, forwarding, and configuration transitions independently of the Flock Block alert log.
Passive and active modes are separate targets. A product detectable while sending Bluetooth scan requests cannot be counted as detected in a passive-only configuration unless another documented signature is present.
Negative and positive controls
Negative controls include consumer access points, phones, laptops, televisions, printers, watches, earbuds, vehicles, tags, and a congested environment with no supported scanner target. Positive controls replay or operate a documented supported signature. Analysts must not tune thresholds against the held-out evaluation runs.
Run procedure
- Synchronize clocks and record the environment, channel use, physical layout, and equipment inventory.
- Run a minimum baseline with Flock Block absent, then present but off.
- Run randomized target-active and target-inactive trials with the operator blinded where practical.
- Repeat across distance, line-of-sight, body obstruction, wall obstruction, antenna orientation, and congestion levels.
- Capture raw scanner observations and Flock Block logs without storing unrelated third-party identifiers.
- Run the decoy function separately from detector evaluation and count only records actually observed by the reference receiver.
- Preserve failed, interrupted, and ambiguous trials with exclusion reasons rather than silently deleting them.
Metrics
Detection reporting includes target-level sensitivity, alert latency, missed detections, false-alert rate per hour, confidence intervals, and results by configuration. Decoy reporting includes unique observed temporary addresses or identifiers, advertisement or frame counts, churn, protocol mix, reference-receiver loss, and the real-device observations that remain present.
No result may be generalized to “all scanners.” Supported-target results must include the tested version. No alert is reported as “no scanner present”; it is reported as “no supported signature observed during this window.”
Publication and change control
Each release will include the protocol version, field schema, analysis code, aggregate results, equipment matrix, exclusions, and limitations. Any protocol change after data collection begins receives a dated changelog and an explanation of whether it was made before analysts viewed outcomes.
Download the machine-readable test field schema. The schema contains no real captures and no customer or bystander data.
Privacy and safety
Tests use owned or authorized equipment in controlled spaces. Raw public data must hash or replace device addresses with per-experiment pseudonyms, omit unrelated traffic, and avoid precise third-party location. Testing must not transmit disruptive traffic, bypass access controls, intercept protected content, or interfere with emergency or licensed communications.
Questions or peer-review notes can be sent to contact@flockblock.app.