Most people think of Google as being the full extent of the
Internet. However, this is a common misconception since Google only
accounts for about 20% of the Internet. The dark web and the deep
web constitute the remainder of the Internet. The dark web is a
section of the Internet where users can access websites
anonymously. Due to its anonymity, the dark web has become
associated with illegal and criminal activities. On the dark web,
access to compromised web cameras through which unsuspecting
victims are watched can be purchased. This has been an ongoing
concern for many who have started taking measures to ensure
security and privacy, such as covering up device cameras out of
fear of being monitored.

The intersection of Artificial Intelligence
(“AI”) and surveillance will be
inevitable as society’s digital footprint continues to
increase and surveillance technology improves. A machine learning
algorithm has recently been released which was trained on security
camera footage in order to predict a person’s bank PIN code.
The model uses security camera footage to predict a person’s
bank PIN code based on how the person covers the keypad when
entering their PIN at an ATM machine. This shows how AI algorithms
can be trained on video footage to achieve specific goals or
purposes. In this article, we discuss how artificial intelligence
can be used on a web camera for malicious purposes, such as the
creation of deepfakes.

What are deepfakes?

Deepfakes are artificially intelligent algorithms which
replicate a person’s likeness, such as voice and facial
features. To launch a robust deepfake attack, a hacker will need
access to a sizable amount of data on the victim, including
pictures and videos which will be used to train the deepfake. For
those in the public realm, such as celebrities and athletes, this
data is readily available which is why we have seen an increase in
reliable deepfake replicas. Many will recall the deepfake of Mark
Zuckerberg in 2017 where he stated the following: “Imagine
this for a second: One man, with total control of billions of
people’s stolen data, all their secrets, their lives, their
futures.” Although deepfakes are in their nascent stages,
with the majority of deepfakes being used for Internet trolling or
in jest, it does not mean that deepfakes are harmless.

Deepfakes introduce a number of extreme risks including inter
alia:

• Reputational damage: deepfakes can be used to create fake
  videos or images that damage a person’s reputation if they
  depict the person as engaging in inappropriate, illegal, or
  embarrassing activities;
• Identity theft: deepfakes can be used to impersonate an
  individual, potentially leading to identity theft, fraud, or other
  criminal activities;
• Privacy invasion: deepfakes can invade a person’s privacy
  by superimposing their likeness onto explicit or intimate content
  without their consent. Again, this could also cause reputational
  harm;
• Misinformation: deepfakes can spread false or misleading
  information;
• Blackmail: hackers can use deepfakes to extort money or other
  concessions from individuals by threatening to release fake
  content;
• Trust erosion and zero-trust society: deepfakes can erode
  trust, as people become increasingly sceptical about the
  authenticity of online media; and
• Social engineering: hackers may use deepfakes to deceive others
  into taking actions they otherwise wouldn’t, potentially
  causing harm to the individual or others. For example, a deepfake
  of a person’s voice could be used to authorise a bank
  transfer.

As surveillance and people’s digital footprints increase
in tandem and AI becomes more advanced, it is arguably only a
matter of time before their convergence. This may result in the
creation of more advanced deepfakes which are harder to detect and
identify, leading to higher levels of fraud and identity theft, and
ultimately the erosion of trust resulting in a zero-trust
society.

The bottom line is that the smaller your digital footprint, the
more difficult it will be to create a deepfake of you. This is a
factor which is currently impeding the large-scale adoption of
deepfake technologies. However, this will change if hackers link AI
algorithms to live webcam footage. Ultimately, this could result in
more robust deepfakes which can be used to commit identity theft,
fraud, or defamation.

For protection against these threats, we recommend the
following:

• Review accounts and delete any inactive accounts;
• Regularly update passwords and do not use variations of the
  same password for different accounts;
• Set up a privacy sticker to block webcams when not in use;
• Install and regularly update antivirus and adblockers;
• Avoid clicking on links unless the source can be verified;
• Avoid accessing ‘untrusted’ websites; and
• Avoid downloading any attachments of emails originating from
  unknown senders.

The content of this article is intended to provide a general
guide to the subject matter. Specialist advice should be sought
about your specific circumstances.

POPULAR ARTICLES ON: Technology from South Africa