Device Fingerprinting and Risk Scoring

GrantFlow collects device information during role activation requests to support AI-driven risk scoring. This helps approvers make informed decisions by providing context about the device and environment from which access is being requested.

Purpose

Device fingerprinting serves several security objectives:

  • Anomaly detection: Identify when a request comes from an unusual device or location
  • Risk assessment: Provide signals for AI-based risk scoring algorithms
  • Audit trail: Record device context for compliance and forensic investigations
  • Pattern recognition: Detect potentially compromised accounts based on behavioral changes

What Information Is Collected

When a user submits a role activation request, GrantFlow collects non-invasive device signals from the browser. This information is attached to the activation request and stored as part of the audit trail.

Browser Characteristics

SignalDescriptionExample
User AgentBrowser and OS identificationChrome 120 on Windows 11
PlatformOperating system platformWin32, MacIntel, Linux x86_64
LanguagesBrowser language preferencesen-US, de-DE
TimezoneUser's timezone offsetUTC-5 (EST)
Screen ResolutionDisplay dimensions1920x1080

Hardware Signals

SignalDescriptionPrivacy Note
CPU CoresLogical processor countGeneric hardware profile
Device MemoryApproximate RAM (GB)Rounded to nearest GB
GPU Vendor/RendererGraphics hardware identifierWebGL-derived information
Touch SupportTouch-enabled deviceBoolean flag

Network Information

SignalDescriptionUse Case
Connection TypeNetwork type (wifi, cellular, ethernet)Anomaly detection
Downlink SpeedEstimated bandwidthProfile consistency
Round-Trip TimeNetwork latency estimateLocation inference
IP AddressRequest source IPGeo-location, anomaly detection

Security Signals

SignalDescriptionPurpose
WebDriver DetectionAutomated browser detectionBot prevention
Headless BrowserHeadless mode detectionAutomation detection
Browser IntegrityExtension/modification detectionTampering detection

Privacy Considerations

GrantFlow's device fingerprinting is designed with privacy in mind:

What We Don't Collect

  • No persistent device IDs: We don't create or store cross-session tracking identifiers
  • No cookies for fingerprinting: Device data is collected fresh with each request
  • No third-party tracking: All collection happens client-side without external services
  • No personal files: We never access local storage, files, or personal data

Data Handling

  • Session-scoped: Fingerprint data is associated only with the specific activation request
  • Hashed sensitive data: Font lists and similar high-entropy data are hashed before storage
  • Tenant isolation: Device data is stored within tenant-specific databases
  • Audit purpose: Data is retained as part of the audit trail per your retention policies

Transparency

Users can see that device information is collected as part of the activation request process. The information is used solely for security assessment and audit purposes.

How Risk Scoring Works

Device fingerprint data feeds into GrantFlow's risk scoring system:

Risk Factors

The AI risk scoring model considers factors such as:

  1. Device consistency: Is this request from a device the user has used before?
  2. Location anomaly: Is the apparent location unusual for this user?
  3. Time patterns: Is the request at an unusual time?
  4. Automation signals: Are there signs this isn't a human-initiated request?
  5. Network characteristics: Is the network profile consistent with expectations?

Risk Levels

Risk scores are categorized into levels that approvers can see:

LevelScore RangeGuidance
Low0-30Normal request, standard approval process
Medium31-60Some unusual signals, review carefully
High61-80Multiple anomalies, verify with requester
Critical81-100Strong anomaly indicators, consider denying

INFO

Risk scores are advisory. Approvers always have final authority over approval decisions, and should use their judgment based on context they may have that the automated system doesn't.

Viewing Device Information

For Approvers

When reviewing an activation request, approvers can see:

  • Summary risk score and level
  • Key device characteristics
  • Notable anomalies flagged by the system
  • Comparison with the user's typical request patterns

For Administrators

Administrators have access to:

  • Full device context in audit logs
  • Historical device patterns per user
  • Risk score distribution analytics
  • False positive/negative feedback for model improvement

Browser Compatibility

Device fingerprinting works across all modern browsers:

BrowserSupport LevelNotes
Chrome/EdgeFullAll signals available
FirefoxFullAll signals available
SafariPartialSome hardware signals limited by privacy features
Mobile BrowsersFullTouch and mobile-specific signals included

TIP

Safari's Intelligent Tracking Prevention may limit some signals. This doesn't affect core functionality—the risk scoring model adapts to available signals.

Compliance Considerations

GDPR

Device fingerprinting for security purposes typically falls under "legitimate interest" for protecting systems and data. However, you should:

  • Include device data collection in your privacy policy
  • Ensure retention periods align with your data protection policies
  • Document the security purpose in your records of processing activities

SOC 2

Device fingerprinting supports SOC 2 compliance by:

  • Providing additional authentication context
  • Creating detailed audit trails
  • Enabling anomaly detection controls

Industry Regulations

For regulated industries (finance, healthcare), device fingerprinting provides additional assurance that access requests originate from expected sources. Consult your compliance team for specific guidance.

Enhanced Fingerprinting Signals

Recent updates have expanded the fingerprinting signal set to improve device identification accuracy:

SignalDescriptionAdded
WebGL RendererGPU and graphics driver identification via WebGL rendering contextv1.3
Canvas FingerprintUnique rendering characteristics of the HTML5 Canvas APIv1.3
Audio ContextAudio processing characteristics unique to hardware configurationv1.3
Font EnumerationAvailable system fonts detected via rendering measurementv1.2
Hardware ConcurrencyNumber of logical CPU cores reported by the browserv1.2

These signals work together with existing fingerprinting data (timezone, screen resolution, language, platform) to create a more reliable device identifier. All signals are collected client-side and hashed before transmission — no raw fingerprint data is sent to the server.

INFO

WebGL rendering identification may not be available in all browsers. When WebGL is disabled or blocked, GrantFlow falls back to other available signals. This does not affect functionality but may reduce fingerprint uniqueness.

Troubleshooting

Missing Device Data

If activation requests lack device information:

  1. Check browser compatibility: Ensure the user's browser supports the required APIs
  2. Review network proxies: Some corporate proxies strip client information
  3. Check for browser extensions: Privacy extensions may block fingerprinting APIs

High False Positive Rate

If too many legitimate requests are flagged as high risk:

  1. Review threshold settings: Consider raising thresholds if they're too sensitive
  2. Provide feedback: Use the approval interface to mark false positives
  3. Check user patterns: Frequent travelers or users with multiple devices may naturally have more variation

Performance Concerns

Device fingerprinting adds minimal overhead:

  • Collection takes less than 50ms client-side
  • No blocking operations in the request path
  • Fingerprint data is small (approximately 2KB per request)