Combating Deepfakes: A Report on AI, Technology, and the Future

This report examines the growing threat of deepfakes, which are AI-generated videos that can convincingly depict individuals saying or doing things they never did. It highlights the increasing prevalence of deepfakes and the potential for misuse, including identity theft and the spread of misinformation. The report details how organizations like Facebook, Twitter, and Google are responding to this challenge, including the use of machine learning and deep learning techniques for detection. It discusses the creation of datasets, the development of AI-based detection tools, and the collaboration among tech companies to combat deepfakes. The report also touches upon the policies and initiatives being implemented to address this evolving technological threat, emphasizing the ongoing need for advancements in AI to stay ahead of the increasingly sophisticated methods used to create deepfakes.

Running head: BATTLE AGAINST DEEPFAKES
BATTLE AGAINST DEEPFAKES
Name of the Student
Name of the University
Author Note

1BATTLE AGAINST DEEPFAKES
Introduction
In June – July (2019), an Amsterdam based cybersecurity company, Deeptrace found
around 14,700 deepfakes. It was an 84 percent increase in deepfakes within six months (since
December 2018). For this reason, many people have faced identity crisis and lost a lot of
money including businesses all around the world. Hence, the post ‘Prepare for a Long Battle
against Deepfakes’ have been chosen to get an idea about what is going on around the world
and for people with no prior knowledge can educate themselves regarding the same.
Discussion
The post is chosen because of its short and detailed analysis about what crisis the
world is facing at the moment (related to deepfake) and what various organizations are
currently doing to minimize rather prevent deepfakes. It portrays people in various videos
where they never appeared before, doing things that they not even once did and saying things
that they never really said in those videos. The effects of these can vary from harmless to
very serious ones. Harmless ones can be superimposed face of actors on their peer’s face
whereas the harmful ones can target famous leading personalities like politicians and people
with high authorities. As a result, the American Senate Intelligence Committee have called in
various technological sharing platform parties like YouTube Facebook and Twitter to
implement new policies against sharing these types of deceitful contents.
In response to this summon, Facebook teamed up with Amazon Web Services,
Microsoft, MIT, along with many other organizations organized a Deepfake Detection
Challenge. The competition officially launched on December 11, 2019 for approximately
three months. Facebook decided on offering a sum of $10 million for the project (Prepare for
a Long Battle against Deepfakes-KDnuggets, 2020). The detection process includes use of
advanced techniques like machine learning and deep learning. Python is mostly preferred for

2BATTLE AGAINST DEEPFAKES
this process; as a result, tools like PyCharm and Jupyter notebook are used. These techniques
require large datasets of real and fake videos to properly train and prepare the machine
learning process. To help researchers develop proper deepfake detection methods, Google has
released a huge deepfake video datasets. Addition to this, datasets like Microsoft’s 1million
dataset, the ‘celebrity together’ dataset and IMDB-WIKI are also available. Apart from
Facebook, Twitter focused on re-constructing its policies to fight against the deepfakes and
was open for taking public inputs in creating deepfake policies.
Conclusion
With the appearance of fake applications, people with no knowledge of programming
skills can also make imitating fake videos. Even though big technological companies like
Google, Microsoft, Amazon and Facebook are fighting their best against the modified videos,
the publicized AI technologies behind the deepfake videos are also getting refined day-by-
day resulting in increase of difficulty in fighting against doctored videos. Due to this reason,
they are planning to come up with AI based deepfake detection tools that will also update
with the deepfake making technologies.

3BATTLE AGAINST DEEPFAKES
Reference
KDnuggets. 2020. Prepare For A Long Battle Against Deepfakes - Kdnuggets. [online]
Available at: <https://www.kdnuggets.com/2020/02/long-battle-against-
deepfakes.html> [Accessed 29 March 2020].