Page 44 - Banking Finance December 2024

ARTICLE

KYC Fraud using Deepfake: Deepfake KYC fraud is a intelligence (AI) and deep learning algorithms. The process
growing concern in India, where fraudsters leverage AI- involves manipulating or replacing a person's voice, style or
generated deepfake videos and images to bypass Know- actions in existing media to make it appear as if they are
Your-Customer (KYC) verification processes. This has saying or doing something they never did. Deepfakes can
significant implications for financial institutions, fintech mimic a person's voice and facial features using artificial
companies, and other organizations that rely on KYC to verify intelligence, It can mimic person’s facial movements from
the identity of their customers. videos or even just a picture of their face and uses an audio
recording of person’s voice to make a realistic looking video
In the recent past, we came across many other deepfakes and sound that the person might never have done or said.
videos & images of celebrities and political leaders whose
alleged video/images were used during the election How deepfake works:
campaign or advertisement or in public forums.
Gathering Data: The technology behind deepfakes,
called "deep learning," needs a lot of pictures or videos
In this article, we will learn and understand about Deepfake, of the person they want to fake. The more data it has,
its impacts in the banking industry and suggestive measures the better it can learn to mimic that person's
to prevent and mitigate such fraud.
appearance and movements.
Deepfake Training the Model: This data is used to train a
computer program to understand how the person's face
A fake/ artificial image or a series of images or video looks from different angles, how they talk, and how they
generated by “deep” learning (a special type of machine
move. It's like teaching the computer to become an
learning) is known as ‘Deepfake’. Simply, we can say that it
impersonator.
is nothing, but a “fake” images or videos generated by
“deep” learning technology. Creating the Fake: Once the computer program has
learned enough, it can then create new videos or images
that look like the real person but are completely fake.
Imagine someone taking a video of you and replacing your
face with someone else's face so convincingly that it looks
like you're saying or doing things you never did. That's a Generally, identifying deepfake videos or images can be
challenging as the technology is constantly evolving.
deepfake.
However, there are several clues and techniques which can
Deepfake technology is a method of creating a real looking be used to spot or identify them:
but fake media—videos, images, or audio—using artificial Unnatural Movements: Pay close attention to facial
expressions, eye movements, and lip-syncing. Deepfakes
often struggle to perfectly replicate natural human
movements, leading to inconsistencies.

Inconsistent Lighting and Shadows: Look for
mismatches in light and shadows and how it falls on the
face compared to the surroundings. Deepfakes often fail
to accurately mimic natural lighting conditions.

Blurring and Artifacts: Examine the edges of the face,
especially around the hairline and jawline. Deepfakes
may exhibit blurring or unnatural smoothness due to the
manipulation process.
Skin Tone and Texture: Check if the skin tone matches
the rest of the body and if the texture appears too
smooth or artificial.
Blink Rate: Deepfakes often have irregular or infrequent
blinking patterns.

BANKING FINANCE | DECEMBER | 2024 | 39