Digital Tampering Detection (E.g., Photoshop)

Key Capabilities

• Image-Editing Software Detection: Identifies receipts that have been edited with image manipulation software (e.g., Photoshop).
• Metadata Analysis: Detects suspicious changes in image or PDF metadata.
• Copy-Paste Detection: Identifies where parts of the receipt image have been copied and pasted.

How It Works

When a file is submitted, our Taggun's engine performs the following analyses:

1. Error Level Analysis (ELA): Scans for inconsistencies in image compression that may signal tampering.
2. Copy-Paste Detection: Flags areas where parts of the image have been duplicated, often used in fraud.
3. Metadata Inspection: Reviews the image or PDF metadata for any signs of editing.

Receipts are flagged for potential tampering should be manually reviewed or rejected.

Setup Process

1. Contact Taggun

Reach out to [email protected] to enable this feature for your account.

2. Use Verbose Endpoints

This feature is available in all verbose API responses. It will be returned automatically once enabled for your account.

📘

Note on Receipt Validation APIs

Digital Tampering Detection is currently only available in Taggun Data Extraction APIs, not in Receipt Validation APIs. For inquiries about using this feature with receipt validation, please contact us at [email protected] to discuss your use case and potential solutions.

Understanding Response Properties

Main Properties

Tampering Detection Details Properties

Examples

1. Image-Editing Software

A user modifies a receipt using Photoshop to change the total amount from $85.20 to $852.00.

Response:

{
  "entities": {
    "tamperDetection": {
      "data": {
        "isTampered": true,
        "tamperedScore": 0.83,
        "details": {
          "elaTampered": true,
          "metadataTampered": true,
          "copyPasteTampered": true,
          "whiteBoxTampered": false,
          "duplicates": [
            {
              "x1": 520,
              "y1": 780,
              "x2": 540,
              "y2": 800
            },
            {
              "x1": 560,
              "y1": 780,
              "x2": 580,
              "y2": 800
            }
          ],
          "rectangles": [
            {
              "x": 500,
              "y": 760,
              "width": 100,
              "height": 60
            }
          ],
          "percentageBlocksTampered": 2.15,
          "normalisedOverallMeanDiff": 28.76,
          "clusterPercentage": 0.89,
          "blocksTampered": 5184
        }
      }
    }
  }
}

2. Copy-and-paste

A user modifies a receipt by screenshotting the "$" symbol and a "5" digit from elsewhere on the receipt and pasting them to change the total amount from $52.00 to $552.00.

Response:

{
  "entities": {
    "tamperDetection": {
      "data": {
        "isTampered": true,
        "tamperedScore": 0.79,
        "details": {
          "elaTampered": false,
          "metadataTampered": false,
          "copyPasteTampered": true,
          "whiteBoxTampered": false,
          "duplicates": [
            {
              "x1": 510,
              "y1": 785,
              "x2": 525,
              "y2": 800
            },
            {
              "x1": 250,
              "y1": 400,
              "x2": 265,
              "y2": 415
            },
            {
              "x1": 530,
              "y1": 785,
              "x2": 545,
              "y2": 800
            },
            {
              "x1": 180,
              "y1": 350,
              "x2": 195,
              "y2": 365
            }
          ],
          "rectangles": [
            {
              "x": 505,
              "y": 780,
              "width": 45,
              "height": 25
            }
          ],
          "percentageBlocksTampered": 1.08,
          "normalisedOverallMeanDiff": 18.32,
          "clusterPercentage": 0.76,
          "blocksTampered": 2592
        }
      }
    }
  }
}

3. Metadata Tampering

A user modifies the metadata of a receipt image to change the creation date, making an old receipt appear to be more recent for a reimbursement claim.

Response:

{
  "entities": {
    "tamperDetection": {
      "data": {
        "isTampered": true,
        "tamperedScore": 0.68,
        "details": {
          "elaTampered": false,
          "metadataTampered": true,
          "copyPasteTampered": false,
          "whiteBoxTampered": false,
          "duplicates": [],
          "rectangles": [],
          "percentageBlocksTampered": 0,
          "normalisedOverallMeanDiff": 0.005,
          "clusterPercentage": 0,
          "blocksTampered": 0
        }
      }
    }
  }
}

Use Cases

• Promotional Campaigns: Validate that receipts haven’t been altered to meet promotion criteria.
• Expense Management: Prevent reimbursement based on tampered receipts.

FAQs

Q: Will this feature catch all types of digital manipulation?

A: While our system is designed to detect a wide range of tampering techniques, no system is 100% foolproof. We continuously update our algorithms to catch new manipulation methods.

Q: How can I differentiate between intentional tampering and innocent image adjustments?

A: Review the tamperedScore and details in the response. Innocent adjustments (like cropping) typically result in lower scores.

Best Practices

1. Human Review: Implement a human review process for receipts flagged as tampered.
2. Comprehensive Protection: Combine with other fraud detection tools for better protection.
3. User Education: Educate users about the consequences of submitting falsified receipts.

Getting Started

Get Started: Email [email protected] to enable tampering detection on your account.