Skip to main content

​
Overview

Fintelite AI’s fraud detection analyzes documents for signs of tampering, manipulation, and authenticity issues. It combines multiple detection techniques to provide comprehensive risk assessment.

​
Detection Categories

Font Analysis

Detects inconsistent fonts, sizes, and styling

Image Analysis

Identifies image manipulation and editing artifacts

Metadata Analysis

Examines document metadata for inconsistencies

Manipulation Detection

Finds signs of text or content alteration

​
How It Works

1

Upload Document

Provide the document to analyze via file upload, URL, or file ID
2

AI Analysis

Multiple AI models analyze the document for fraud indicators
3

Risk Assessment

Results are compiled into a comprehensive risk score
4

Get Report

Receive detailed fraud analysis report with evidence

​
Running Fraud Detection

Submit a document for fraud analysis (async recommended): 
curl -X POST https://api-vision.fintelite.ai/fraud-async \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "files": "id://file-uuid",
    "webhook": {
      "url": "https://your-app.com/webhook"
    }
  }'
Returns job ID immediately: 
{
  "id": "job-uuid",
  "status": "IN_QUEUE"
}
Completed job response structure: 
{
  "id": "job-uuid",
  "type": "FRAUD",
  "status": "COMPLETED",
  "execution_time": 4500,
  "output": {
    "type": "json",
    "output": [
      {
        "anomalies": [...],
        "font_analysis": [...],
        "image_analysis": [...],
        "manipulation_analysis": [...],
        "metadata_analysis": [...],
        "summary": {
          "risk_level": "WARNING",
          "issues_detected": ["Font Inconsistencies"],
          "metrics_summary": "1 of 4 triggered review-level indicators."
        }
      }
    ]
  }
}
For sync vs async processing details and job status management, see the Jobs Concept Guide.

​
Understanding Results

​
Risk Levels

TRUSTED

Document appears authentic with no significant issues detected

WARNING

Any fraud indicators detected, document requires review

HIGH_RISK

High risk fraud indicators detected, document likely fraudulent (Coming Soon)
Currently, the system returns either TRUSTED or WARNING risk levels. HIGH_RISK level is planned for future release.

​
Indicator Status

Each fraud indicator has a status:
  • PASS: No issues detected in this category
  • WARNING: Potential issues that require attention
  • CRITICAL: Clear indicators of manipulation or fraud

​
Analysis Categories

The fraud detection response includes:
  • font_analysis: Font consistency, size variance, and styling checks
  • image_analysis: Image manipulation, JPEG artifacts, and clone detection
  • manipulation_analysis: Text overlay and content alteration indicators
  • metadata_analysis: Document metadata inconsistencies and creation info
  • anomalies: Detailed list of specific issues with severity levels and evidence
  • summary: Overall risk level, issues detected, and metrics summary

​
Use Cases

Identity Verification

Detect tampered ID cards and passports

Financial Documents

Verify authenticity of invoices and bank statements

Legal Documents

Check contracts and agreements for alterations

Insurance Claims

Validate claim documents and receipts

​
Integration Patterns

Fraud detection can be combined with data extraction for comprehensive document verification:
  1. Parallel Processing: Run fraud detection and data extraction simultaneously
  2. Risk-Based Routing:
    • TRUSTED: Auto-approve and process
    • WARNING: Flag for manual review
    • HIGH_RISK: Reject immediately
  3. Multi-Stage Validation: Use fraud results to filter documents before extraction

​
Best Practices

Don’t rely solely on the overall risk level. Review specific anomalies for context, especially those with HIGH or CRITICAL severity. Anomalies provide detailed evidence about what triggered the fraud indicators.
Store fraud detection results for audit trails and compliance. Keep records of risk levels, issues detected, anomalies, and processing timestamps for future reference and dispute resolution.
Use fraud detection alongside other verification methods:
  • Database lookups
  • Third-party verification APIs
  • Business rule validation
  • Historical pattern analysis

​
Limitations

Fraud detection is a powerful tool but not infallible:
  • False Positives: Legitimate documents may trigger warnings
  • Sophisticated Fraud: Advanced forgeries may pass undetected
  • Document Quality: Poor quality scans may cause false alarms
  • Context Matters: Always review results in business context

​
Performance

  • Supported Formats: PDF, JPG, PNG, HEIC, TIFF
  • Page Limit: Up to 10 pages per document
  • Best Results: High-resolution scans (300 DPI+)

​
Next Steps

Document Processing

Learn about data extraction

Fraud API

Explore fraud detection endpoints