NAVIGATION

Metadata analysis provides automated detection of document integrity by extracting and analyzing potential inconsistencies, unauthorized modifications, or manipulations hidden within the metadata of the uploaded files across both image and PDF formats. Ultimately, its primary objective is to reinforce trust and verification within document submission workflows, exposing hidden file anomalies and fraud risks that typically evade manual inspection.

1. Objectives

Current Limitations: Document review processes were prone to digital file manipulations, causing compliance risks.
The Opportunity: The Metadata Analysis feature offers automated integrity checks that surface technical anomalies, ensuring document authenticity without manual intervention.
Target Persona: Compliance Officers, Credit Officers, and Document Processors are the user types most impacted by this feature.

2. Feature grasp

A technical breakdown of the metadata feature.

Component	Description	Impact
Field/Metadata parameters	New granular document metadata extraction fields	Enables tracking of file modifications and origin.
Logic/Workflow	Automated integrity verification during document ingestion	Instantly identifies anomalies to provide clear insight into the file's origin
Integration	Direct API capabilities supporting document metadata analysis responses	Allows seamless deployment into existing document submission workflows.

Key User Story

As a Compliance Officer, I want to automatically verify the integrity of uploaded documents, so that I can reduce fraud risks and ensure document authenticity before any financial review.

3. Practical Use Cases

Real-world scenarios demonstrating the feature's value.

Scenario A: Bank statement Processing

Context: An officer uploads a PDF Bank statement for processing.
Application: The feature automatically detects inconsistency in the document's modification timeline and software signatures.
Outcome: The submission file is flagged for forensic review, successfully preventing potential financial fraud.

Scenario B: National ID Processing

Context: An officer submits a digitally modified Identification image.
Application: The feature identifies a metadata mismatch between the file’s hardware & software information and the internal editing history.
Outcome: The system prevents potential identity fraud, protecting the organization as well as the customer.

4. Step-by-Step Instructions

When enabled for your organization, metadata analysis is seamlessly integrated into the submission process. It automatically evaluates the metadata of all ingested files without impacting the processing speed of your existing workflow.

5. Understanding the metadata

PDF metadata analysis
Parameter name	Result explanation
PDF Producer	The underlying software engine, library, or application that actually generated or converted the document into the PDF format.
Modification Date	The exact date and time the PDF file was last edited, saved, or otherwise altered after its initial creation.. If this is much later than the creation date, the file may have been edited.
Font Usage	A technical list of all the typefaces (fonts) used in the document. Unusual font patterns can indicate that text was edited or added after the document was created.
PDF Creator	The original software application used to author the document before it was converted into a PDF. This helps identify whether the file came from a legitimate source or a document editing tool.
Creation Date	The original date and time the PDF file was generated. This helps compare the file creation time with other dates in the document.
Title	The formal, human-readable name of the document stored inside the metadata. A missing or unusual title is not always suspicious, but it can help understand the file source.

Image metadata analysis
Parameter name	Result explanation
Resolution Unit	The unit of measurement used for the image's pixel density. Missing or unusual values can indicate the image was processed using an external tool or converted.
Orientation	Tells the computer how the camera was held. Missing or unusual values can happen when an image is edited, exported, or stripped of metadata.
DateTimeOriginal	The exact date and local time the photograph was originally taken by the camera. Native photos often include this, while screenshots or edited files may not.
ModifyDate	The date and time the image file was last edited, saved, or altered. This helps detect whether the image was changed after capture.
SubsecTimeOriginal	This records the exact fraction of a second the shutter clicked. Native camera photos include this, while screenshots and processed images often do not.
GPS	The geographic location and time where the photo was taken. Missing GPS is common and not always suspicious, but it helps with authenticity checks.
Software	The name and version of the software or camera firmware used to create, edit, or process the image. If an editor, scanner, or converter appears here, the image may have been processed.
Image Description	A text field intended for a brief title, caption, or description of what the image shows. Sometimes screenshots, generated images, or edited files store clues here.
User Comment	Checks hidden user comment metadata. Some devices or apps store screenshots or processing information here.
PNG Filter Method	Checks PNG mathematical pre-processing method used to prepare the image data. Certain values can suggest the image is a screenshot or was converted from another format.
Make	The manufacturer or brand of the camera or device used to take the photo. Native photos often contain this information, while screenshots or edited images may not.
Model	The specific model name or number of that camera or device. Missing or generic model information may indicate the image was not taken directly from a camera.
Lens Model	The exact lens attached to the camera when the photo was captured. Native photos may contain this, while screenshots or processed images usually do not.
ICC Profile Copyright	The legal copyright text for the color profile embedded in the image. This can help identify whether the image came from a real device, screenshot, editor, or conversion tool.
JUMD/JUMBF label	Checks for C2PA or JUMBF labels, which may indicate AI-generated, edited, or content-credential metadata.

6. Preview

</aside>