> ## Documentation Index
> Fetch the complete documentation index at: https://docs.veridox.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Document Analysis

> Master the lifecycle of AI-powered forensic document verification.

## Overview

Veridox provides comprehensive forensic document analysis through a suite of AI-powered detection engines. This guide covers the end-to-end analysis lifecycle, from initial case creation to interpreting detailed fraud indicators and risk assessments.

## The Analysis Lifecycle

Document verification in Veridox follows a structured path:

1. **Case Creation**: Group related documents into a single container for context-aware analysis.
2. **Document Ingestion**: Securely upload files to our storage infrastructure.
3. **Automated Processing**: Veridox engines perform forensic analysis in parallel.
4. **Intelligence Retrieval**: Access granular findings, authenticity scores, and generated reports.

<Info>
  **Supported Formats**: Veridox is optimised for images (JPEG, PNG, TIFF) and documents (PDF), analysing them for metadata anomalies and pixel-level tampering.
</Info>

## Forensic Capabilities

### Image Forensics

<AccordionGroup>
  <Accordion title="Metadata & Provenance Analysis">
    Validates <Tooltip tip="Exchangeable Image File Format">EXIF</Tooltip> data to ensure digital integrity:

    * Verifies capture device signatures and software markers.
    * Validates capture timestamps and GPS coordinates.
    * Detects signs of metadata stripping or manipulation.
  </Accordion>

  <Accordion title="Error Level Analysis (ELA)">
    Detects compression inconsistencies within image layers:

    * Identifies spliced regions with varying compression ratios.
    * Highlights localised editing or pixel-level manipulation.
    * Generates a visual heatmap for investigation.
  </Accordion>

  <Accordion title="Noise & Filter Analysis">
    Examines sensor noise patterns to detect synthetic elements:

    * Identifies inconsistent digital noise across different regions.
    * Detects cloning, "healing," and copy-paste operations.
    * Flags AI-generated or enhanced regions.
  </Accordion>

  <Accordion title="Visual Search & Stock Detection">
    Performs global reverse image searches:

    * Identifies the use of public stock photography.
    * Detects reused images across different verification cases.
    * Matches content against known fraudulent document repositories.
  </Accordion>

  <Accordion title="Content Authenticity (C2PA)">
    Validates cryptographic provenance markers:

    * Verifies hardware-level digital signatures.
    * Tracks the immutable edit history of the file.
    * Validates compliance with the C2PA standard.
  </Accordion>
</AccordionGroup>

### Document Forensics (PDF)

<AccordionGroup>
  <Accordion title="Structural Integrity">
    Analyses the internal architecture of the PDF:

    * Detects anomalous creation tools or non-standard libraries.
    * Flags modification markers and hidden object streams.
    * Validates PDF versioning and cross-reference tables.
  </Accordion>

  <Accordion title="Formatting & Font Consistency">
    Examines visual and structural consistency:

    * Identifies font substitution and "invisible" text overlays.
    * Detects layout anomalies and alignment inconsistencies.
    * Validates barcodes and security patterns against global standards.
  </Accordion>

  <Accordion title="Entity Extraction & OCR">
    Transforms documents into structured data:

    * High-accuracy OCR for text extraction.
    * Automatic recognition of persons, dates, and reference numbers.
    * Cross-validation between extracted text and embedded metadata.
  </Accordion>
</AccordionGroup>

## Interpreting Risk Levels

Every document is assigned a risk classification based on the aggregate findings of all analysis engines:

| Risk Level        | Description                                             | Recommended Logic                               |
| :---------------- | :------------------------------------------------------ | :---------------------------------------------- |
| **Informational** | No fraud indicators. Standard characteristics detected. | Proceed with automated acceptance.              |
| **Low**           | Minor anomalies found that are likely benign.           | Accept with optional background monitoring.     |
| **Medium**        | Moderate risk factors or multiple minor anomalies.      | Route to a human agent for manual review.       |
| **High**          | Significant indicators of tampering or fabrication.     | Reject automatically or escalate to fraud team. |

## Highlights

Analysis results include a `highlights` field — a plain list of key findings summarising the most significant indicators detected:

```json theme={null}
{
  "highlights": [
    "Metadata manipulation detected",
    "AI-generated content indicators found"
  ]
}
```

This is useful for quickly surfacing the most critical signals without parsing the full `findings` array. The field is omitted when the analysis did not produce any highlights.

## Detailed Findings

Each analysis report includes a `findings` array. A typical finding follows this structure:

```json theme={null}
{
  "heading": "Anomalous Document Reference Format",
  "category": "Document Integrity",
  "description": "The reference number uses a 24-digit format instead of the 11-digit standard.",
  "supporting_evidence": [
    "Standard DVLA V5C format requires 11 digits.",
    "The embedded Code128 barcode matches the 24-digit string."
  ]
}
```

## Real-Time Notifications

You can use [webhooks](/guides/webhooks) to receive automatic notifications when file analysis completes, rather than polling for results. This enables automated workflows like risk-based escalation and report generation.

## Security & Best Practices

* **File Quality**: High-resolution uploads (min. 1200x1600 pixels) ensure the highest accuracy for ELA and noise analysis.
* **Access Control**: You can only access and retrieve analysis results for documents you have personally created or requested.
* **Data Retention**: Upon account termination, Veridox retains files for 90 days after permanent deletion. Ensure you download the [PDF reports](../api-reference/endpoints/files-analysis-get) if you require long-term storage for compliance.