Page 125 - Cyber Defense eMagazine September 2025

The Technology Behind Deepfakes

Deepfakes are primarily created using machine learning algorithms, especially GANs. A GAN consists of
two neural networks, the generator and the discriminator, that work together to create data
indistinguishable from real-world examples. The generator crafts synthetic media, while the discriminator
evaluates their authenticity, prompting iterative improvements in realism. As a result, GANs can produce
highly convincing images, videos, and voices.

The accessibility of deepfake creation tools has expanded dramatically. Open-source libraries and
commercial platforms now allow individuals with modest technical skills to produce deepfakes. Video and
audio impersonation can be achieved with relatively small datasets, sometimes just a few minutes of
video or audio of the target. The rise of deepfake-as-a-service platforms means attackers can now
outsource the creation of synthetic content, lowering barriers for cybercriminals and increasing the
proliferation of malicious synthetic media.

Real-world Threats and High-profile Incidents

Deepfakes have already played a role in several documented cyberattacks. In 2019, a deepfake audio
attack successfully impersonated a CEO’s voice to trick a subordinate into transferring €220,000 to a
fraudulent account. Similar incidents have targeted companies, government agencies, and individuals,
with attackers using synthetic voices or faces to bypass security controls or manipulate victims.

Cybercriminals are increasingly leveraging deepfakes for spear-phishing campaigns and business email
compromise (BEC) scams. Adversaries can use AI-generated audio or video to impersonate executives
or trusted officials, lending authenticity to fraudulent requests. Political deepfakes have been deployed to
spread misinformation or undermine public trust during elections, fueling disinformation campaigns that
can influence opinions and destabilize societies.

Moreover, deepfakes are used for extortion, harassment, and blackmail. Malicious parties can fabricate
compromising images or videos, threatening personal reputations and causing emotional and
psychological harm. As the technology advances, it becomes more challenging for victims, and even
experts, to distinguish real from fake, escalating the overall risk landscape.

Impact on Digital Trust and Society

At an individual level, deepfakes pose severe risks to privacy, reputation, and financial security. Synthetic
media can be used for identity theft, impersonation, and personal attacks that leave lasting damage on
victims’ lives. For organizations, deepfakes open the door to social engineering attacks, corporate
espionage, fraud, and significant brand damage. A convincing deepfake can trick employees, investors,
or customers, resulting in both direct financial loss and long-term erosion of trust.

Democracies face even broader implications. Deepfakes can spread false information on social media,
fueling disinformation campaigns that erode trust in public institutions and the media. Fabricated videos

Cyber Defense eMagazine – September 2025 Edition 125
Copyright © 2025, Cyber Defense Magazine. All rights reserved worldwide.