Fraud Detection 2026 – Trends in Theses  
Part 2: Metadata Analysis

Let’s talk about metadata—in Part 2 of our series on fraud and fraud detection trends (if you missed Part 1: link in the first comment).

We are sticking to the thesis format, but becoming more specific: every single digital document contains metadata—and it is a treasure trove for fraud detection. Those who know how to analyse it are already on the trail of manipulation. Why?

How does fraud detection software approach metadata analysis?

1. Metadata from image files includes creation date, software used, editing timestamps, and often even the location where the image was captured.

2. Medical invoices are rarely created using tools such as GIMP or Photoshop, but rather with dedicated practice software. If image editing software or an AI image generation platform appears in the metadata of submitted “original” documents, this should raise suspicion.

3. Discrepancies between creation date and invoice date can also indicate irregularities, as documents properly created by legitimate authors typically show consistent data.

4. Camera data such as smartphone model or GPS coordinates can provide indications of authenticity as a positive criterion.  
What else?

5. PDF metadata also enables analysis of fonts used, layers, and the separation of individual elements to identify manipulation and compare them with genuine documents from the same originating system.

6. Metadata can be removed or altered—clever, these fraudsters. However, the absence of relevant information is often another warning sign.

7. Whether using tools such as GIMP, IrfanView, Windows Explorer, Master PDF Editor, Microsoft Edge, a purchased full fake from the darknet, or a prompted AI-generated image: there are always metadata traces…  

8. …and they are almost never identical to the “correct” metadata patterns of the legitimate author.  
What does this mean overall?

9. Fraud detection software such as ICO.Fraud is capable of extracting metadata directly from files and transmission logs and converting it into structured data, enabling ICO-LUX GmbH’s solutions to provide valuable insights to fraud investigation units.

10. This information can then be used for AI-supported pattern recognition, plausibility checks, and risk assessments—significantly reducing the chances of fraudsters succeeding.

Intriguing, isn’t it?

And it stays intriguing: in the next article, we will share insights into image analysis.

Speaking of images: our cover image was manually prompted by ICO-LUX. All it takes is a good idea and a well-crafted prompt—the rest is now handled by NanoBananaPro.