Getting started
Guides
Concepts
Resources
Doc X-Ray is a powerful machine-learning technology that estimates the authenticity and correctness of your clients' PDF documents.
Trained in-house on tens of thousands of documents, classified and processed in six different languages, this machine-learning model can categorize documents with an accuracy of 93%.
Documents can be uploaded either during the creation of a new request, from the Documents tab of an existing case or from an API request. Doc X-Ray can read an exhaustive list of documents, including financial statements, bank statements, bank details, registration documents, and ID cards.
To extract information from a document, we use Optical Character Recognition (OCR) technology. However, OCR technology only works on "True PDF" documents. A True PDF is a digital document converted into a PDF format using a software. On True PDF, the text is separated from images and both the meta-data and the characters in the text hold an electronic character designation.
Once you have successfully uploaded your client documentation, our OCR technology will immediately begin analyzing it. The classification and display of the detailed report will be available within a few minutes after the upload.
Doc X-Ray aims to detect signs of fraud attempts by analyzing the documentation uploaded with a financing case. Doc X-Ray analyzes the file metadata structure and captures data from a document, looking for patterns in structure, font, and labeling, and finally identifying signs of editing.
An overall flag alert is displayed alongside a more detailed analysis, which allows you to investigate each sign of mischief further, as identified by our technology.
Our technology performs 11 checks on uploaded documents to detect potential fraud attempts. Doc X-Ray raises an alert every time a suspicious element is detected. These alerts focus on:
Below are details about the alerts and their level of severity:
| Alert name | Definition | Alert category | Severity |
|---|---|---|---|
| Edited text | Parts of the text have been modified and differ from the text original version. These modifications are highlighted in the document using red boxes. | Content | Fraudulent |
| Inconsistent font | Consecutive letters or numbers have different fonts and sizes compared to the rest of the document. Font inconsistencies are highlighted in the document using yellow boxes. This alert can indicate either a fraud attempt or a mistake in the document. | Content | Moderate |
| Overlaid & overlapped text | Block (with or without text) covering an existing text element. In that case, the original text is still visible on the PDF metadata. These modifications are highlighted in the document using yellow boxes. | Content | Moderate |
| Purposefully altered document | Multiple versions of a single document have been detected. By clicking on this alert, you can access the Doc comparison tab to compare the original version of the document and the latest version found. | Document structure | Fraudulent |
| Encrypted document | Document protected from modifications. Encrypted documents cannot be analysed by our OCR technology. | Document structure | Moderate |
| Questionnable creator tool | The document has been created using a tool allowing modifications on the document, such as Word or Adobe Indesign. The “Creator” can be found in the Doc Info tab. | Document structure | Moderate |
| Edited by a suspicious software | The document has been edited using a software allowing modifications on the document, such as Adobe Acrobat. | Document structure | Moderate |
| Different creation and modification dates | Every document has a digital signature which includes the creation date and last modification date of a document. If these dates aren’t the same, this means that the document has been modified after its generation. | Document structure | Moderate |
| Different generation and upload times | This check calculates the time separating the creation of the document and its upload on October Connect, as a short period of time may indicate potential modifications. | Document structure | Moderate |
| Fingerprint | To prevent fraudsters who may manage to get around all our checks, we compare the document to a set of similar documents to detect unusual elements. This is only available on French documents. | Document structure | Moderate |
| Document integrity | Every PDF source code has a summary table which indicates the structure of the PDF. If the PDF structure doesn’t match with the summary table, the document has been altered. | Document structure | Moderate |
A Doc X-Ray flag is generated as soon as the last uploaded document is analyzed. It can be classified as:
The Doc X-Ray Feedback Loop has been enhanced to give users greater control over fraud detection. With this update, users can now:
Access to this feature is securely managed and can be enabled or disabled by the administrator, ensuring it is available only to the appropriate team members.
A score is calculated at the case level for every time you create a new case from your October Connect account. This score is generated based on all the alerts raised in all the documents belonging to the case.
The calculation of this overall score is based on the weight and the severity of every potential alerts that could be detected by Doc X-Ray on any document.
Once we get this value, we compare it with the score obtained by SMEs from similar sector and geography. Depending on how well or bad the company does compared to their peers, we transform this score into a letter going from A to E (overall positive to overall negative).
← Previous
Next →