Behavioral Biometrics

v2.1.0 · Human behavior pattern analysis generating ~165 bits of unique entropy from mouse movements, keyboard typing rhythm, scroll behavior, and touch interactions.

Privacy note: The engine does not record actual keystrokes — it only captures timing patterns (key hold duration, inter-key intervals). Single characters are anonymized as "char"; only special keys (Shift, Enter, etc.) retain their names.

Public API — `window.__behavioralV2`

Method	Returns	Description
`start()`	`void`	Begin 5-second data collection
`stop()`	`void`	Stop collection early
`getData()`	`Object`	Get raw behavioral data (mouse, keyboard, scroll, touch)
`generateFingerprint(data)`	`string`	8-character behavioral hash from collected data
`hasEnoughData()`	`boolean`	Whether minimums are met for fingerprinting
`getState()`	`Object`	Current collection state (counts, timing)

Collection Configuration

Parameter	Value	Description
`collectionDuration`	5,000ms	Auto-stop after 5 seconds
`maxEventsPerType`	100	Cap per event type to prevent payload overflow
`minMouseEvents`	10	Minimum mouse events for valid fingerprint
`minKeyEvents`	5	Minimum keyboard events
`minScrollEvents`	3	Minimum scroll events

Collected Data Points

Mouse Events

Field	Type	Source
`x`, `y`	number	`clientX` / `clientY`
`movementX`, `movementY`	number	Pointer delta
`timestamp`	number	`Date.now()`
`type`	string	`"move"` \| `"click"`
`button`	number	Click button index (clicks only)

Keyboard Events

Field	Type	Source
`key`	string	`"char"` (anonymized) or special key name
`duration`	number	Key hold time (keydown → keyup) in ms
`timestamp`	number	`Date.now()` at keyup

Scroll Events

Field	Type	Source
`scrollY`	number	Absolute scroll position
`deltaY`	number	Scroll delta per event
`timestamp`	number	`Date.now()`

Touch Events

Field	Type	Source
`x`, `y`	number	Touch position
`force`	number	Touch pressure (0–1, device-dependent)
`type`	string	`"start"` \| `"move"`
`timestamp`	number	`Date.now()`

Pattern Analysis Metrics

The tracking engine (tracking.js) performs advanced analysis on the raw data before sending it to the backend:

Mouse Metrics

{
  avgVelocity: 0.85,          // px/ms average cursor speed
  velocityVariance: 0.23,     // Speed consistency
  avgAngleChange: 0.42,       // Radians avg direction change
  clickCount: 3,
  moveCount: 47,
  straightLineRatio: 0.15     // % of near-straight movements
  // Bot indicator: straightLineRatio > 0.8
}

Keyboard Metrics

{
  avgInterval: 185.3,         // ms between keystrokes
  intervalVariance: 2450.1,   // Typing rhythm consistency
  avgKeyDuration: 72.5,       // ms key hold time
  durationVariance: 890.2,
  keyCount: 23,
  suspiciouslyRegular: false  // true if variance < 10 (bot)
}

Scroll Metrics

{
  avgDelta: 53.2,             // Average scroll delta
  deltaVariance: 1205.8,
  avgInterval: 312.5,         // ms between scroll events
  intervalVariance: 8920.3,
  eventCount: 8,
  uniformDelta: false         // true if all deltas identical (bot)
}

Fingerprint Generation

The behavioral fingerprint is an 8-character hex hash derived from:

signature = [
  mouseVelocity.toFixed(2),
  keyboardRhythm.toFixed(2),
  scrollPattern.toFixed(2),
  touchSignature.toFixed(2),
  mouseEventCount,
  keyboardEventCount,
  scrollEventCount,
  touchEventCount
].join('|');

hash = simpleHash(signature) → 8-char hex

Integration with Tracking Engine

// Automatic flow (triggered by tracking.js):
1. tracking.js sends main event (t=0s)
2. tracking.js calls __behavioralV2.start() (t=0s)
3. Behavioral engine collects for 5s (t=0-5s)
4. tracking.js reads data at t=6s
5. If hasEnoughData() → POST /t with behavioral payload

// The behavioral payload is sent as a SEPARATE request
// with action: "behavioral_analysis"

Backward Compatibility

Legacy globals are maintained for older integrations:

window.__trackBehavioral = startCollection;     // Legacy alias
window.__getBehavioralData = getBehavioralData;  // Legacy alias
window.__behavioralData → getBehavioralData();   // Getter (copy)