Fraud Detection Tools - Overview

Key Features

Taggun offers a comprehensive suite of fraud detection and prevention tools:

1. Duplicate and Similarity Check

   • Identifies identical or highly similar receipts to prevent double submissions.
   • Detects slight modifications made to receipts in an attempt to reuse them.
   • Uses advanced algorithms to compare new submissions against historical data.

2. Digital Tampering Detection

   • Recognises digitally altered receipts, including changes to dates, amounts, or merchant information.
   • Uses advanced image analysis to spot inconsistencies and manipulations.
   • Employs multiple detection methods, including metadata analysis and pixel-level examination.

3. Handwritten Receipt Detection

   • Identifies handwritten receipts, which can be more susceptible to fraud.
   • Allows for separate handling or additional verification of handwritten receipts.

How It Works

1. Submit Receipt: When you send a file to Taggun APIs, it can be passed through our fraud detection engine.

2. Analysis: Our system analyses the file using various fraud detection algorithms.

3. Scoring: Each file receives a fraud score based on the analysis.

4. Results: The API returns the fraud detection results along with the extracted receipt or invoice data.

Sample API Response

Here's an example of what the fraud detection section in the API response might look like:

{
  "similarReceipts": [
    {
      "referenceId": "REC-20240105-ABC123",
      "userId": "USER-9876543",
      "trackingId": "T-20241005-9276439",
      "similarityScore": 0.9779201745986938
    },
    {
      "referenceId": "REC-20240106-XYZ789",
      "userId": "USER-1234567",
      "trackingId": "T-20241005-2921841",
      "similarityScore": 0.9779201745986938
    }
  ],
  "tamperDetection": {
    "data": {
      "isTampered": false,
      "tamperedScore": 0,
      "details": {
        "elaTampered": false,
        "metadataTampered": false,
        "copyPasteTampered": false,
        "whiteBoxTampered": false,
        "duplicates": [],
        "rectangles": [],
        "percentageBlocksTampered": 0.7817460317460317,
        "normalisedOverallMeanDiff": 25.143336197326164,
        "clusterPercentage": 0.9370182365106223,
        "blocksTampered": 37233
      }
    }
  },
  "handwritingDetection": {
    "data": {
      "isHandwritten": false,
      "handwrittenScore": 0.05660377358490566
    }
  }
}

In this example:

• similarReceipts shows two receipts that are very similar to the submitted one, with high similarity scores.
• tamperDetection indicates that no tampering was detected (isTampered: false), with detailed metrics on various tampering checks.
• handwritingDetection suggests this is not a handwritten receipt (isHandwritten: false), with a low handwriting score.

Leveraging Taggun's Data for Enhanced Fraud Detection

While Taggun provides powerful fraud detection capabilities, you can further enhance your fraud prevention strategy by leveraging the extracted data. Here are some additional measures you can implement:

Anomaly Detection

• Identify unusual patterns in receipt submissions, such as abnormal amounts or frequencies.
• Flag receipts that deviate significantly from established norms for further review.
• Implement machine learning models to adapt to changing patterns over time.

Geographical Analysis

• Detect suspicious patterns based on the locations of purchases.
• Flag unlikely sequences of transactions across different locations within short time frames.
• Compare transaction locations with known user locations or travel patterns.

Merchant Verification

• Cross-reference merchant information with trusted databases to verify legitimacy.
• Flag transactions from merchants with a history of fraudulent activity.
• Implement additional checks for high-risk merchant categories.

Time-based Analysis

• Detect suspicious patterns in the timing of receipt submissions.
• Flag receipts with dates far in the past or future.
• Identify unusual patterns in the time gap between purchase date and submission date.

Amount Distribution Analysis

• Analyse the distribution of transaction amounts to identify outliers.
• Implement category-specific amount thresholds based on historical data.
• Flag transactions that are just below approval thresholds or company policies.

User Profiling

• Build profiles of normal user behavior and flag significant deviations.
• Implement risk scoring based on user history and characteristics.
• Use machine learning to adapt user profiles over time.

Best Practises

1. Customisation: Tailor fraud detection thresholds to your risk tolerance. What's suspicious or risky for one situation might not be for another.
2. Comprehensive Approach: Use a combination of all available fraud detection features for the most robust protection. Complement Taggun's receipt analysis with additional fraud prevention mechanisms, such as:
   1. User behavior analysis
   2. IP address monitoring
   3. Device fingerprinting
   4. Two-factor authentication for high-risk transactions
3. Human Review: Implement a process for human review of flagged receipts to minimise false positives and negatives. This ensures good customer service and support by:
   1. Training reviewers on common fraud patterns and your specific business rules
   2. Establishing clear escalation procedures for suspicious cases
   3. Regularly auditing the review process to maintain consistency and effectiveness
   4. Providing a clear and simple appeal process for users whose submissions are rejected
4. Regular Updates: Keep your fraud detection settings up-to-date to combat evolving fraud techniques. Consider periodic reviews of your fraud detection thresholds and rules.
5. User Education: Educate your users about proper receipt submission to reduce unintentional flags. This can include:
   1. Clear guidelines on acceptable receipt formats and quality
   2. Tips on taking clear photos of receipts
   3. Information on what constitutes fraudulent behavior and its consequences
   4. Regular reminders and in-app guidance on best practices

Conclusion

By combining Taggun's fraud detection tools with these additional strategies, you can create a robust, multi-layered approach to fraud prevention that significantly reduces your risk exposure while maintaining a smooth user experience for legitimate transactions.