Thumbprint™ by MoneyThumb contains patent pending AI technology that analyzes whether a bank statement has been altered or is fraudulent.
Lenders routinely request bank statements to verify income, cash flow, or assets. However PDF copies of bank statements can be altered or even completely fabricated. Doing a google search on 'create a fake statement' comes up with dozens of sites that will create a completely fictitious bank statement with whatever amounts you want. Or one can start with a genuine bank statement and use PDF editing tools such as Adobe Acrobat Pro and change the text in a bank statement, to alter numbers, payees, and totals, to reflect greater income and or cash flows.
Typical methods for authenticating bank statements are first to check for obvious mistakes such as typos or inconsistencies in typefaces. Then look for unusual formatting or text that’s seen to be out of position compared to the rest of the document. You can also compare the document against a known genuine document to look for formatting differences.
Check the financial institution name and logo on the statement. Verify that the column setup and fonts match those from a known authenticate statement.
One method to detect fake bank statements is to reconcile the totals. That is to total up all the deposits, withdrawals, checks, and fees and see if the totals match the balances printed on the statement. Every bank prints some kind of totals, whether its starting/ending balances, a running balance, or both. A high percentage of fraudulent statement don't pass this test, simply because people make mistakes when they edit the numbers.
Authenticating third party documents such as bank statements is not a new problem. Experienced underwriters can recognize many subtle discrepancies in format and thereby detect many fraudulent statements. However, new software tools make it easier than ever to create fictitious bank statements, that are harder than ever to detect.
MoneyThumb created thumbprints for thousands of banks by aggregating data from processing millions of PDF bank statements. A bank Thumbprint consists of hundreds of data points that go into creating a bank statement. For example, what columns are used, what date formats, what text fonts, the exact position of each column, and how transaction amounts are encoded in the PDF file itself. Experienced underwriters already know to look for many of these characteristics, but others are either hard to spot, or just not visible to the human eye. MoneyThumb bank thumbprints contains numerous attributes that can only be found with a computer program. Finally, machine learning is used to evaluate whether a bank statement matches the expected Thumbprint.
As an analogy, consider formatting a document in an editor such as Microsoft Word or Google Docs. You can add white space by using spaces, tabs, or indents. It makes no difference, and is indistinguishable to the reader which one you used. However, if you always use spaces, and then someone else edits the documents and uses indents, its clear that you were not the one that made the final version of the document. That is the type of validation that Thumbprint does for PDF documents.
How to determine if a PDF document has been altered?
Thumbprint™ from MoneyThumb does two types of document verification. The first is to look for internal inconsistencies. For example, is all the text in a given column the same font, size, and at the same position. And position is not just position visible to the human eye, its position measured to an accuracy of 1 ten-thousands of an inch. Additional thumbprint factors are all evaluated using an algorithm created by machine learning.
The second type of document authentication is to compare the document thumbprint to the thumbprint of other statements from the same bank. All the factors used in internal consistency checks are evaluated along with additional factors such as the software program used to create the PDF document.
The combined result is a score from 1 to 1000, where 1 indicates no evidence of alteration, and 1000 is a document that has almost certainly been altered. And its all done within just a few seconds!
Every document has a Thumbprint.
