Divergence experiment

Browser Identity Lab

FingerprintJS plus passive rendering, capability, media, storage-privacy, execution, and cross-realm probes. Collection requests no permissions or user interaction. A protected first-party cookie and a non-exportable installation key provide automatic continuity evidence.

Measurement label Loading FingerprintJS…

Overall evidence posture

Trust overview

Waiting for measurement

— %

Evidence coverage —

Raw risk load —

Integrity retained —

Assurance —

Enforcement disabled

Evidence coverage × integrity retention = overall trust

Raw risk contributions remain uncapped and auditable. Integrity retention uses 100% ÷ (1 + raw risk load ÷ 100), avoiding a hard zero when evidence exceeds 100%.

Evidence family	Availability	Coverage	Explanation

Passive WebGL scene

Particle field v1

Waiting for measurement

The scene measures one deterministic frame automatically, then continues as a non-measuring live visual. Its pixel digest is render-state evidence; live animation remains diagnostic and non-scoring.

Sample duration —

Frame rate —

Frame p90 —

Frame render —

Dropped estimate —

State digest —

24,576 deterministic particles

Cross-channel evidence

Environment integrity

Waiting for measurement

Raw risk load —

Detected conflicts —

Challenge state —

Recent interruptions —

Combined server-side evidence from browser realms, rendering APIs, workers, challenge lifecycle, and focus activity. Individual weak signals never prove automation.

Signal	Risk contribution	Strength	Evidence

Advanced probes

Hardware and execution signals

Versioned passive evidence is collected automatically. Timing samples remain non-scoring and are exported separately for analysis.

Probe	Status	Identity summary	Metrics	Duration

Server component digest —

Client FPJS confidence —

Server-observed IP —

Changed components —

Post-measurement workload

GPU benchmark v2

Waiting for measurement

Render index —

Compute index —

Max variability —

Execution —

GPU cluster —

Performance cluster —

Claimed model —

Identity consistency —

Runs after the observation is saved, normally in a Web Worker. Version-specific throughput is diagnostic only and excluded from device matching and scoring.

Render tier	Iterations	Median batch	Draws / second	M iteration-pixels / second	Variability

Current measurement

User-agent profile

Waiting for measurement

—

No user agent collected

—

Observed sources

HTTP request header is compared with the main window, dedicated/shared workers, and sandboxed iframe.

Server-observed evidence

HTTP and Client Hints profile

Waiting for measurement

Edge source —

HTTP protocol —

UA-CH fields —

Header names —

Captured by the server on challenge and observation requests. Cookie and authorization values are never retained; missing headers are not treated as fraud.

Request	HTTP User-Agent	Sec-CH-UA-*	Sec-Fetch-*	Other allowlisted headers	Trusted edge

Experimental evidence

Browser automation

Waiting for measurement

Evidence points —

Tool artifacts —

WebDriver —

Detected signals —

Client/server parity —

The browser and server run the same versioned classifier. Only the server result is authoritative; disagreement is separate integrity evidence.

Signal	Result	Strength	Evidence

Persistent storage

Measurement history

Each row is compared with the immediately preceding measurement.

#	Label	Client	Server digest	Client visitor ID	Client confidence	Server IP	Integrity	Changed	Time

Current measurement

Fingerprint components

Duration is diagnostic only and is excluded from the server-side comparison.

Component	Value or error	Duration