Fraud in Images: The Value of Human Analysis in Due Diligence Investigations
Tuesday, September 11, 2018
the breakneck speed of development in artificial intelligence and the
increasing availability of large databases of corporate information, the
instincts of an experienced human investigator can uncover “red flags” based
upon hunches that aren’t available to machine algorithms.
conducted standard due diligence research for a client on a fuel importer in a
Latin American country, including confirmation of the company’s registration, details
on any legal filings or criminal history for the company, a review of sanctions
lists, and more. Part of the research involved a review of the company’s
website. Such a review typically locates information on the history of the
company, its business activities, and its clients, but generally doesn’t raise
any “red flags”. Company websites are marketing tools and unlikely to
purposefully reveal information harmful to the company’s reputation.
materials can, however, misrepresent a company’s true business capabilities,
and the incentive exists for companies to make it appear as if they own and
operate more productive facilities or are in a stronger financial position than
they might actually be.
case, misrepresentation existed in background images on the website rather than
the text of a press release, brochure, or presentation. The company’s website
had what appeared at first glance to be a standard slideshow of images at the
top of the site showcasing scenes relevant to their line of work: a tractor trailer
with the company’s logo; a tank car; and fuel storage silos — each with the
company’s logo on them. These images gave the impression that the company owns the
equipment needed to do the job and is an established, prosperous entity.
these images just didn’t “look right.” Something subtle in the lighting or the
angle of the logo raised our investigator’s suspicions. As presented, the
images rotated too fast for a close inspection, but through digging into the
site’s publicly available code, we were able to isolate the specific
The most telling “photo” was of a fuel
storage terminal. Each tank had an image of the company’s logo shown on it, and
while it didn’t “feel right”, we need something we could point to that would
more definitively illustrate our inkling. We found it on the second tank: clear
evidence of the image having been manipulated. In the world of this image, there
is a light pole towards the right of the scene that stands vertically in front
of the tank, but where the image shows the words “fuel supplier” below the
company’s logo, the “r” appears on top of the light pole rather than behind it,
as it would if those words were painted on the tank.
This and the other images give the viewer the
impression that the company owns and operates the tanks and vehicles shown, yet
the manipulation apparent in one of these images should rightly give one pause
in thinking this to be true.
else should the client question? Has the company truthfully answered the
client’s questionnaire? What could be
discovered if a visit was made to the company’s location? What information
could be unveiled by making discreet inquiries with the company’s neighbors or
others in the industry?
a trained skeptical eye, a viewer would likely skip over these images and not
give them a second glance. These images should activate a heightened sense of
skepticism regarding the company’s self-representation; it is a sort of key
that can predicate the direction of an investigation. These feelings and
observations are unavailable to algorithmic and AI-centric due diligence
reviews and highlight the value of the human investigator.