Impact Of Deepfake Technology

I have given the formatting rules as well.

please add figures and tables in results please.just need to add around 20 pages in the results chapter.

including figures, tables and references.

Implications of Deepfake Technology on Individual Privacy and Security
2
Directions for the Paper writing:
1. Please completely avoid using surveying tools in the project to avoid the required
approvals. This project was initially made for conducting survey but now I must exclude the
survey and make this paper purely conceptual based on current trends on
awareness levels on Deepfakes and its implications and emphasize on the
2.
3.
4.
5.
6.
fact that there is huge need for development of rules and regulations, awareness, and detection
mechanisms in deepfake technology.
I have already added conclusion and future work and recommendations and use it to frame the
results chapter accordingly. Please try to add the data in each chapter including Methodology,
Results and Conclusion wherever I have written Need to add more content if it is in RED font
color to summarize the results in emphasizing on the fact that there is huge need for
development of rules and regulations, awareness and detection mechanisms in deepfake
technology based on our research using Literature review.
Please add figures / tables (5 each) in the RESULTS chapter on current trends on implications of
deepfake, awareness levels on deepfakes, potential misuses, prevention mechanisms, reactions
to deepfakes.
Please note that all the side headings and the chapters mentioned need to be present in the
document. The main structure of the document is already provided in the index.
Please add 20 + pages (15 pages of results including tables & figures) from your side for this
effort , apart from the existing 49 pages.
ABSOLUTELY NO PLAGIARISM.
Formatting of paper:
•
•
•
•
•
Font 12pt Arial throughout the document.
Visible page numbers start on the second page (often the abstract page) and that page should
be numbered as 2.
Page margins are 1” (one-inch) on all sides—top, bottom, and sides
The abstract has a bolded title and the body of the abstract is single-spaced and not indented.
The Table of Contents (TOC) is properly formatted
o Section title (Table of Contents) is bolded
o The TOC should be double spaced with no additional spacing in between headings/lines
o If you have tables/figures in your paper, include a line for the “List of Tables” and “List of
Figures” and their respective page numbers in the TOC
o The headings “Chapter” and “Page” are placed above their respective sections of the
table of contents and continue over onto each page of the TOC
o The dot leaders go all the way to the page numbers, which are flush right
3
o
•
•
•
•
•
•
•
•
•
•
•
•
•
All chapter titles as well as level 1 and level 2 headings must be included in the Table of
Contents. Level 3 headings and beyond are optional in the TOC
List of Tables and List of Figures are properly formatted
o The formatting rules are the same as the table of contents
o All figures/tables should be accounted for
o Titles in the list of figures/tables should match the titles in the paper. Your project must
follow the requirements of one style guide (APA, MLA, etc.) throughout your paper.
APA7 for References
Check that chapter titles are formatted correctly and matching the chapter titles in the table of
contents.
o Each chapter should begin on its own, separate page.
Spacing in between paragraphs follows the style guide used by the author.
Ensure headings and subheadings are formatted correctly and in the correct position on the
page based on style guide requirements. Do not underline or bold the subheadings. Can make it
Italicised.
Citations must be formatted correctly based on the style guide selected and must match their
corresponding reference perfectly.
o Every single in-text citation must have a corresponding reference in the
References/Bibliography page. Use the search or Control F feature to search in-text citations.
o There must be correct year, correct spelling of the author’s name, correct authors,
correct use of et al., etc.
Tables and figures are formatted correctly according to the style guide used.
o Correct location of figure/table number and title.
o Figure/table in correct position on the page (centered or to the left).
o Figures should be cited if they were taken from another source
References/Works Cited are formatted correctly and should all have matching citations
o Every single reference must have a corresponding in-text citation in the paper. Use the
search or Control F feature to search through every reference.
o References/bibliography page must be correctly formatted based on style guide
requirements for each type of source.
Appendices, if included, are formatted correctly.
o If you do have appendices, they should also be listed at the end of your table of
contents
o Format appendix titles based on style guide requirements or as Appendix A: Title of
Appendix (title is centered and bold.
Line spacing is 2 without extra spacing. (Double Spacing)
In references include authors full last name and initial (Follow APA7)
Format the references as per APA 7 requirements. Include full information including all
author’s last names.
4
Tables/Figures:
–
Per APA 7, Table/Figure titles should be located above your tables/figures and should
include the following:
o First line is “Table” or “Figure” and its number (bolded and flush left)
o Second line is the table/figure title which should be in italics and in title case.
o Double-space before and after tables/figures.
o Format any table/figure notes as Note (italicized) followed by a period (e.g. Note.)
Example of correct format:
In-text Citations
–
Ensure that all of your in-text citations are formatted correctly per APA 7, with authors’
last names and date of publication.
–
Please double check that all in-text citations in your paper all have a corresponding
reference entry
5
Abstract
Technological advancements not only can make life easier but also create
menacing consequences when misused. Deepfake technology, being one of the major
advancements recently, serves as an example of such technology. It is extremely
difficult to identify the fake media from the real media. Deepfake technology uses
Artificial Intelligence to impersonate someone and create hyper-realistic media like
videos and pictures. This study investigates the origination of deepfake technology, its
impact on our society and business, existing and future regulations that might be
needed, measures to detect, mitigate, and prevent the deepfake misuses. These
findings underscore the necessity and significance of adopting a multifaceted strategy to
address deepfake risks, which should include prevention, regulations, education, and
public awareness efforts.
•
Change this abstract to conceptual Research paper for starred paper as a study
on Implications of Deepfake Technology on Individual Privacy and Security.
6
Table of Contents
Abstract …………………………………………………………………………………………………………………………………….. 5
Acknowledgements …………………………………………………………………………. Error! Bookmark not defined.
List of Tables …………………………………………………………………………………………………………………………….. 8
List of Figures …………………………………………………………………………………………………………………………… 9
Chapter I: Introduction …………………………………………………………………………………………………………….. 10
Introduction………………………………………………………………………………………………………………………….. 10
Problem Statements ……………………………………………………………………………………………………………. 11
Nature and Significance of the Problem ………………………………………………………………………………. 11
Study Questions ………………………………………………………………………………………………………………….. 12
Limitations of the Study ……………………………………………………………………………………………………….. 12
Definition of Terms ………………………………………………………………………………………………………………. 12
Chapter II: Background and Review of Literature ……………………………………………………………………. 14
Introduction………………………………………………………………………………………………………………………….. 14
Background Related to the Problem ……………………………………………………………………………………. 14
Literature Related to the Problem ………………………………………………………………………………………… 19
Summary …………………………………………………………………………………………………………………………….. 36
Chapter III: Methodology …………………………………………………………………………………………………………. 37
Introduction………………………………………………………………………………………………………………………….. 37
7
Design of the Study……………………………………………………………………………………………………………… 37
•
Data Collection …………………………………………………………………………………………………………. 37
•
Data Analysis ……………………………………………………………………………………………………………. 38
•
Data Presentation: Need to add more content …………………………………………………………. 38
Summary …………………………………………………………………………………………………………………………….. 38
Chapter IV: Results …………………………………………………………………………………………………………………. 39
Introduction………………………………………………………………………………………………………………………….. 39
Results and Analysis: ………………………………………………………………………………………………………….. 39
Impact of Deepfakes on Democracy and Public Trust among the Society …………………………… 40
Summary …………………………………………………………………………………………………………………………….. 41
Chapter V: Conclusion and Recommendations ………………………………………………………………………. 43
Conclusion …………………………………………………………………………………………………………………………… 43
Recommendations ………………………………………………………………………………………………………………. 44
Future Work…………………………………………………………………………………………………………………………. 45
References ……………………………………………………………………………………………………………………………… 47
8
List of Tables
9
List of Figures
Figure
Page
10
Chapter I: Introduction
Introduction
Information technology has improved tremendously over the last decade.
Unfortunately, misuses of technology are also on the rise. One such misuse is deepfake
media which has been negatively impacting society and their discourse.
Deepfake is an artificial intelligence technology, which can create hyper-realistic
media such as images and video. This made it possible to create audio or video of a real
person saying and doing things he or she never said or did. Deepfake can be highly
deceiving and dangerous as it has a high potential to manipulate the public’s opinions
and their decision making. It may also create havoc in victims’ lives. While the deepfake
initially targeted on political leaders, celebrities, and artists, it may extend its misuse
among ordinary people. For example, it can be used in creating porn videos as a bullying,
revenge, and extortion tool. As technological advancement is inevitable and so are their
threats to people if misused, it is highly essential to become aware of such technology
and to create a proper plan to address the issues involved with it.
This paper presents an overview of the emergence of Deepfake technologies, their
benefits and usage, means of detection, societal impact, and measures to mitigate
misuse. It delves into existing legislation, regulations, and policies aimed at regulating
such incidents. It discusses training and education on deepfake technology and recent
technological developments to detect deepfake media. The remainder of this chapter
includes sections on problem statements, the nature and significance of the study,
limitations, and definitions of some key terms.
11
Problem Statements
1. Lack of Public Understanding: There is a significant lack of public understanding
regarding deepfake technology, its capabilities, and its potential risks. This gap in
knowledge hampers effective public action and decision-making.
2. Insufficient Empirical Research: Current literature has little empirical studies that
investigate public perception and awareness of the potential harms associated
with deepfake technology. This absence of data hinders the development of
targeted educational and preventive measures.
3. Legislative Actions: There is a scarcity of research capturing public opinions on
governmental legislative actions aimed at reducing the risk of deepfake
technology. Timely public opinions can help lawmakers in the creation of policies
that are both effective and publicly supported.
Nature and Significance of the Problem
There is a great need for people to understand the risk of deepfake technology. By
addressing the gap in public understanding of deepfake technology, the government and
educational institutions can make better policy decisions. This is important for democratic
societies where public opinion can influence policy decisions. Understanding the public’s
awareness level can also help technology companies to develop effective deepfake
detection tools, making technology a part of the solution. By understanding and
addressing the potential harms of deepfake technology, the study contributes to
maintaining social trust and it can be eroded by malicious uses of the technology. Finally,
the study can provide a basis for educational programs and training materials to counter
the impact of potential misuse.
12
Study Questions
•
What is deepfake, and how did it emerge?
•
Where is deepfake technology currently used?
•
What are the benefits and disadvantages of deepfake?
•
What are potential harms caused by misuse of deepfake technology?
•
To what degree is society aware of the Deepfakes?
•
What are the mitigative measures available currently to alleviate the damage
caused by misuse of deepfakes?
•
What are the available regulations to curb the misuse of deepfakes?
Limitations of the Study
As a novel technology, the literature on the impacts of deepfakes is limited, and
empirical research on its psychological effects is scarce. Furthermore, it was challenging
to find specific cases of deepfake misuse due to confidentiality and legal restrictions,
particularly in cases of deepfake pornography, where actual incidents may not be
accurately reported due to the defamatory nature.
Definition of Terms
Deepfake: A video or an image of people in which either their face or body are digitally
modified to make them appear as someone else.
Fake news: It is a type of journalism which runs on with an aim to deliberately spread
misinformation and false media on news platforms or social media.
Artificial intelligence: It is the ability of a computer to do tasks that are usually done by
human beings where human intelligence is required.
13
Supercharging scams: The scams in which deepfake audio is used to impersonate the
person on the other line is a higher-up such as a CEO and soliciting an employee to send
money.
GPT-2: Generative Pretrained Transformer-2- is an AI model which can predict the token
in the upcoming sequence in an unsupervised way. GPT-2 is a text generating AI released
by the research lab named OpenAI.
GAN: GAN stands for Generative Adversarial Networks which is one of the latest
advances in the deep learning technologies which deals mainly with image recognition,
data computation, and broader analysis which involves activities like assessing and
recapitulating the main four macro-environmental factors such as political, economic,
socio demographic and technological that are the basis of the main changes that take
place in the world. (CVISIONLAB, n.d.)
RNN: Recurrent neural network – type of artificial neural network which uses sequential
data or time series data. Long Short-Term Memory is a deep learning architecture used
in RNN.
Markov Chain: Markov Chain is a mathematical system that experiences transitions from
one state to another according to certain probabilistic rules.
14
Chapter II: Background and Review of Literature
Introduction
In this chapter, we provide background information on deepfake technology,
discuss related incidents, conduct a literature review addressing the problem, and outline
the methodologies used to combat these issues. The literature review includes statistics
and figures illustrating the damage caused by the misuse of deepfake content.
Additionally, we delineate the methodologies based on existing works on deepfake
technology.
Background Related to the Problem
Deepfake technology is developed using artificial images and audio files that are
consolidated along with machine-learning algorithms with an intention to manipulate the
media to create false information. Deepfakes can cause high risks, such as undermining
cybersecurity and influencing political elections. They can also impact the financial
conditions of corporations and individuals, tarnish the reputation of individuals,
organizations, and communities, and lead to disruptions in the lives of individuals in
several ways.
Initially, deepfake’s focus was only on celebrities. However, now-a-days, the
deepfake technology is easily available to the ordinary people and they are able to
develop their own deepfake content and due to this, issues arise such as trying to
manipulate the society and public, invasion of personal space and can attack rights of
individuals, extortion scams, financial frauds and Supercharging scams.
As reported by a visual threat intelligence company named Sensity Systems Inc,
(Petkauskas, 2021), the various deepfake videos that are generated by the AI powered
15
GAN (Generative adversarial Network) have shown overwhelming rise in the reputation
attacks. Not long ago, a report named ‘The State of Deepfakes 2020’ had published that
more than 85 thousand subversive deepfake videos have been created by the expert
designers which were tracked down until December 2020. Since the year 2018, it has
been found that the volume of deepfake videos that are generated by the expert crafters
has doubled for every six months and these deepfake videos were certified as the videos
that were used to either cause harm to the luminaries or potential enough to do so and
this report excluded the list of videos that perform attacks on the individuals.
As mentioned by Giorgio Patrini, CEO, and cofounder of Sensity, there is currently
a great expansion of existing communities of deepfake technology developers and
corresponding content creators. Simultaneously, novel deepfake communities are
emerging globally. Patrini, mentioned in one of his interviews with CyberNews that
pornographic content and deprecatory, derogatory videos that cause reputation attacks
and character assassination, shares a major part of 93% in the list of existing deepfake
videos and also it is the West of the United States that has been a major target in terms
of the misusing technology in attacking Celebrities.
Apart from the reputation attacks on celebrities, the ordinary people are also facing
issues of being targeted by the personal attacks performed using deepfakes. In one of
the reports published in Fall 2020, by Sensity, it is found that a bot network on the
Telegram platform is developed to take photos of women from their social media accounts
and manipulated to stripped off clothing using AI technology which if observed shows how
much havoc it can cause in their daily lives. It is found that over 100,000 women were
victimized by male offenders.
16
One of the best live examples on how deepfake can cause tremors in the lives of
people globally is the deepfake video that had been created on then President of USA,
Donald Trump talking offensively on Belgium’s climate policy (BURCHARD, 2018). This
video was created by one of the local Political Party in Belgium named Socialistische
Partij Anders shortly known as Sp.a and they posted it on twitter and Facebook. It has
created a great aggravation globally and provoked the people to add hundreds of
comments voicing their outrage on then American President that he would dare to
interfere with the Belgium’s Climate Policy and had hate comments on American culture
as well.
However, later the SPA political party confessed that it was done by a team
commissioned by the political party to use machine learning to produce a deepfake video
and posted it with an intention to initiate a public debate to attract attention to the climate
change act. They also claimed that the video was not intended to deceive supporters but
to bring attention to the issue of climate change and said they checked it is legally
acceptable to do so.
Although, the Belgian political party claimed that it is legally acceptable to deliver
a video campaign with the deepfake video, it has created considerable amount of
disturbance across the countries and triggered an unnecessary chaos among public and
increased hate on Americans which is not acceptable. However, this situation shows that
deepfake is not enough to materialize as a threat to democracy and this shows there is a
need to incorporate stringent rules on usage of deepfake technology in the upcoming
future.
17
As mentioned earlier, deepfake is a phenomenon which is a combination of Deep
Learning technology and fake media that are developed using artificial intelligence
technology. It is found that it takes just 300 images of certain person who is a victim of
deepfake to develop a reasonably convincing deepfake media using a swap technique in
which faces of the source and the target are swapped effectively so that it seems hyperrealistic. Irrespective of the technic being employed for the creation of deepfake the
process comprises of three basic steps such as extraction, training and creation which
makes it much easier even without having huge data and a single image of a source will
be enough to create a deepfake in near future. Having said that, there exists various
issues related to deepfakes, Betül Çolak has emphasized legal issues like Intellectual
Property rights and personal data protection in the article “Legal Issues of Deepfakes.”
(Çolak, 2021).
He reported that, according to WIPO (World Intellectual Property Organization) as
published in “Draft Issues Paper on Intellectual Property Policy and Artificial Intelligence”
essentially two questions addressed in it.
1) As the deepfakes are created using the content, subjected to copyrights, who
should be given the copyrights of the deepfake? The source or the creator?
2) Is there a necessity to have a system of equitable remuneration for persons
whose likeliness and “performances” which are used in deepfake? (WIPO, 2019).
WIPO indicates that the deepfake content has immense potential to cause severe
issues such as infringement of human rights, privacy rights, personal data protection right
and so on and so forth. Hence, the WIPO claims that the main concern should be if the
copyrights to be even consented to the deepfake media rather than the concern of whom
18
should be given the copyrights. In response to this concern, WIPO asserts that if the
deepfake content significantly diverges from the victim’s identity, it should not be granted
copyright protection. In cases where copyright applies, it should be attributed to the
creator of the deepfake, given that there is no involvement of the source person whose
image or other media is used in the creation of the deepfake, and it is done with their
consent. It also indicates that copyright is not an appropriate weapon as the victim of
deepfakes do not possess an interest in copyright of their own image. However, the victim
can claim the right of personal data protection to overcome this issue of unethical use of
deepfake.
According to the Betül Çolak, a lawyer specializing in IP and Technology law from
Turkey and a Researcher in the AI and Fairness research cycle of the Institute, both the
states in USA and the BigTech is actively involved in taking action against the issues of
the deepfake content. A best example of this is the active development of the detection
tools for identifying the challenges faced by the deepfake content by the BigTechs and
the introduction of regulations in the states of Virginia, Texas and California, which are
one of the first states to do that in which the law states that there exists criminal penalties
on the circulation of the non-consenting deepfake pornography in the state of Virginia
whereas in the state of Texas, the law forbids the people to create and distribute the
deepfake media that contemplates to inflict upon the candidates of public office or
intended to manipulate the election results. Considering the current issues pertaining to
deepfakes, there is a great need for developing the technological tools and stringent
regulations against deepfake in order to prevent as well as cease the destructive
outcomes of the misuse of deepfakes.
19
Literature Related to the Problem
According to a report on effectiveness of deepfakes in manipulating the attitudes
and intentions of the people generated by conducting experiments on people by exposing
them to the genuine and deepfake media and measured their explicit (self-reported) and
implicit (unintentional) attitudes as well as behavioral intentions and the results obtained
from this experiment indicates that the deepfake media such as video, audio and images
have a severe psychological impact on the audience i.e., viewers and it is just as effective
as the genuine content has in manipulating their attitudes and intentions. (Hughes et al.,
2021). In their study, they have conducted seven preregistered experiments with an aim
to find out what can happen if a viewer happens to come across deepfake content and
what will be the impact of being exposed to deepfake content just for once and how it
affects their biased thinking compared to the actual legit content.
The results suggested that the deepfake video content creators can easily
manipulate the attitude and thinking of the people who are exposed to it thereby making
them vulnerable to giving up control to these deepfake content creators which is a kind of
violation of personal right to think and act rightly as the information that is being put
forward is fake and hence the actions would also be wrong. This study highlighted the
most dangerous aspect of deepfakes, their capability to undermine our beliefs in what is
reality and what is the reliable information that we can trust. ‘Liar’s Dividend’ is a concept
that suggests that some people misuse deepfake technology with the intention of profiting
from the data environment, which is deluged with fake information.”. (Chesney & Citron,
2018)
20
Deepfake technology had originated initially in the computer vision field and later,
subsequently found its usage in audio manipulation and then the text generation (Bregler
et al., 1997). In the recent times, there has been lot of advancements in software which
are capable enough to generate the speech audio text body by manipulating the voice of
various speakers just after listening for five seconds (Jia et al., 2018). According to a
group of deepfake text detection is highly essential as there exists development of
exceptionally advanced generative methods such as GPT-2, RNN, LSTM, Markov chain.
However, it is found in their study that there are not enough methods or software available
to detect the deepfake texts that are visible in huge numbers on social media texts yet.
Also, their study found the existence of 25,572 tweets, comprised of half human and half
bots crafted and posted on twitter. (Fagni et al., 2021).
According to Fred Eslami (Zeman, 2021), an associate at AM Best, a credit rating
agency, says in a world of insurers, the cyber incidents are unique in nature and there is
not much steadfast historical or factual data that can be utilized to predict the
depredations in cyberattacks unlike the natural disasters where the factual and historical
data exists and can be used to predict and prevent the losses. Unlike the regular cyberattacks, deepfakes use artificial intelligence to misrepresent the recorded audio and video
content. Although, initially deepfakes target was movies and entertainment purposes like
humor, eventually deepfakes were used in changing the factual data and spreading the
misinformation which poses the deepfake content a highly threatening tool and the social
media will aid in making them even more dangerous as it is the medium which spreads
the deepfake content instantly across the globe.
21
According to Cybercube, a Saas Company, cyberattacks are increasing at a high
rate using social engineering and various other cyber procedures and when the deepfake
technology meets these techniques it will only raise the success rate of these attacks
exponentially high (Zeman, 2021). One such incident took place in March 2019, where
the cybercriminals used AI based software to mimic the voice of Chief Executive Officer
of a UK based energy firm with an intention to impersonate him to gain unethical profit of
fraudulent transfer of the €220,000 ($243,000) which was identified by cyber-analysts as
an unusual case of AI usage in hacking activities. The cybercriminals were successful in
their attempt to transfer amount to a Hungarian bank account and then distributed to other
locations and due to this the investigators could not trace back the hackers and it caused
a huge loss the company which is extremely unfair to the victims. Many times, cyberattacks cannot be traced easily and may not even recover the losses in these cases.
(Stupp, 2019).
Deepfakes and the Law: A Review of Current Legal Approaches and Regulations:
Deepfakes have become a prominent issue today due to their potential to deceive
and manipulate individuals through the synthetic alteration of audiovisual content using
AI technologies. These sophisticated digital fabrications can blur the boundaries between
truth and falsehood, presenting significant challenges across multiple domains such as
politics, privacy, security, and trust. Recognizing the potential risks and harms associated
with deepfakes, policymakers and legal experts have been engaged in a complex and
evolving discourse regarding the development of effective laws and regulations to tackle
this emerging technology.
22
At present, there is a lack of comprehensive legislation worldwide, specifically
targeting the issue of deepfakes. Each country has its own set of laws and regulations,
leading to variations in approaches. However, it is evident that most countries have not
established dedicated legislation specifically aimed at combating deepfakes. Instead,
existing privacy laws and data protection regulations are often utilized to address
instances of deepfake misuse.
According to Scott Briscoe, a Content Development Director at ASIS International
in his article on Laws addressing the deepfakes, there are stipulations included in U.S.
National Defense Authorization Act (NDAA) to address the growing problem of deepfakes
recently (Briscoe, 2023). According to 2021 NDAA, a law has been created to issue and
annual report on deepfakes by Department of Homeland Security and it should
encompass a comprehensive examination of potential risks and damages associated with
the technology, as well as addressing a wide range of concerns like foreign influence
campaigns, fraudulent activities and harm inflicted upon specific group of people. This
has aided in broadening the scope of deepfake report mandate that was called previously
by NDAA.
According to a report published by Hogan Lovells (Lovells, 2020), a Global law
firm, the European countries like UK, France or Germany has no explicit legislation in
place that directly addresses the legal framework pertaining to deepfakes. In case of any
online disinformation including misuse of deepfakes, the European mandate intends to
address these issues through a set of measures like self-regulatory Code of Practice on
Disinformation for online platforms. This code aims to ensure these online services have
23
security measures against disinformation and checks on availability of appropriate tools
to report the disinformation.
Currently as there is no legislature to combat the misuse of deepfakes present in
European countries, they utilize existing laws, such as laws on deepfakes and the right
to protection from derogatory treatment, to address the deepfake situation as a
workaround.
In case of United states, according to Hogan Lovells report, various states have
recently enacted laws aimed at addressing the detrimental effects of deepfakes.
However, these laws undergo substantial scrutiny by the First Amendment rights of free
speech, and it is uncertain whether courts will determine if these state-level regulations
violate Constitutional principles. Some of the laws that have been created to combat
misuse of deepfakes as listed below:
•
Couple of California laws made effective in 2020 which regulated the distribution
of deepfakes including altered images, audio, or any visual deceptions of a
reasonable individual (Tremaine, 2019). The two laws are described as
o AB 730 – restricting the usage of deepfakes in Political campaigns
manipulation and it comes with an expiry date on January 1, 2023.
o AB 602 – aims to tackle the issues regarding deepfakes and pornography
and do not have any expiration date.
•
Virginia state is one of the first states to introduce laws banning and criminalizing
the unlawful dissemination of falsely created material like digitally generated
pornography known as deepfakes. (Virginia Legislative Information System, 2019)
24
•
Texas state, in September 2019, criminalized the misuse of deepfakes through the
Texas Senate Bill 751 (SB751) amendment to the state’s election code. This law
is enacted only in a political context. The act prohibits individuals to create and
dissemination of deepfakes with an intent to harm a political candidate or
manipulate the outcome of an election (Artz, 2019).
•
Maryland has proposed a bill to legalize prohibition of the misuse of deepfakes
similar to the laws of California in political context. (Lovells, 2020)
According to an article published by Shannon Reid, a graduate from University of
Pennsylvania Law school, since its creation, deepfakes have always challenged the
efficacy of U.S. Law in holding individuals accountable for their actions when they publish
deepfakes of others without their consent. Unfortunately, there are no sufficient solutions
for the targeted individuals in deepfake media in the U.S. federal and state laws. Privacy
laws are insufficient in addressing the specific technologies and behaviors that pose the
most significant threats. Furthermore, federal criminal and intellectual property statutes,
which could potentially apply to deepfakes, are often narrowly interpreted by the courts
or vulnerable to defenses that severely limit the legal options available to potential victims
(Reid, 2021).
Current strategies to detect the deepfakes misuse:
At present, being a novel technology, the presence of standalone detection mechanisms
exclusively designed to combat the misuse of deepfake technology are negligible. A
comprehensive examination of the existing literature reveals a significant gap in the
development of specialized tools and techniques solely dedicated to identifying and
countering the potential malevolent applications of deepfakes. While research in the field
25
of deepfake detection has made significant strides in detecting manipulated content, the
focus has primarily revolved around identifying deepfakes in general without a specific
emphasis on addressing their potential misuse. As the landscape of digital deception
continues to evolve, it becomes increasingly imperative for researchers and technologists
to direct their efforts towards the creation of dedicated measures that can effectively
target the misuse of deepfakes, thereby bolstering the security and trustworthiness of
media content in our digitally interconnected world.
Currently available methods to mitigate the Deepfakes misuse:
The methods to mitigate deepfake misuse have been continually evolving as the
deepfake landscape advances, and a multi-faceted approach is often necessary to
effectively address the challenges posed by deepfake technology.
One of the most popular social networking sites, Meta (formerly Facebook) acknowledges
the severe consequences of non-consensual sharing of intimate images (NCII), often
referred to as “revenge porn.” The company emphasizes its commitment to not allowing
such content on its platforms and announces its efforts to combat the spread of NCII.
Meta and Facebook Ireland, in collaboration with the UK Revenge Porn Helpline and
more than fifty global organizations, have launched StopNCII.org. This unique platform is
designed to offer a secure and confidential resource for individuals who are worried about
the non-consensual sharing of their intimate images, which may include nudity or sexual
content created using advanced AI technologies like deepfake to overcome and avoid
sextortion (Antigone, 2021). The StopNCII.org website is designed to enable the
individuals worldwide to take proactive measures to prevent and stop the unauthorized
sharing of their private images on platforms like social media and other tech websites that
26
are included in this initiative and there by provides greater control to the victims and
enhances the security of their personal media.
There are numerous studies, reports, and research papers on detecting deepfake
technology in the current situation of the cyberworld. A report generated on finding the
effectiveness of deepfake on human attitude and intentions by conducting seven
preregistered experiments in which groups of people have been exposed to the deepfake
videos and genuine videos and tabulated the results on the levels of effectiveness of
deepfake content on a viewer’s intentions and attitude. In their first experiment, they
exposed the viewers to the genuine content which is a video of a novel individual in which
he introduced himself and said few words about highly positive aspects and in another
video, he spoke highly negative statements. A group of viewers watched the negative or
positive videos and then completed corresponding implicit association test and the results
from those tests showed that the genuine online content has strongly influenced the
attitudes of the viewers and their intentions towards the person who spoke in the videos.
This showed that genuine content can promote social learning at implicit and explicit
levels.
In the second experiment, another group of people are exposed to the deepfake
content, and it has been found that the change in attitude and intentions are same as the
genuine content which is a concern to notice. In the fourth and sixth experiments they
tried a different test data in which the group of people exposed to a deepfake content
which was created from scratch and not from the prerecorded media which has been a
successful attempt to control the viewers attitude and intentions.
27
In the third and fifth experiment the group of people acting as viewers are informed
that they will be either exposed to genuine or deepfake content and yet their intentions
and attitudes were changed even after being aware of the presence of deepfake. It means
that deepfake has done its job even when it was an informed act and due to this it makes
a deepfake technology highly dangerous as the damage it caused cannot be undone
easily.
The paper published by (Westerlund, 2019) , highlights the possible threats of the
deepfake technology which include major threats to the political and business system as
well as the society involved with them. It makes the job stressful to the journalists to
identify the real news from the fake ones and they may even give rise to the hard time
trusting them. The study stated that there is a need for legislation and regulation to
encounter the deepfake misuse and it suggests that the deepfakes are remarkable threat
to the society, political and business systems and needs to be combatted by developing
the deepfake detection systems and develop techniques to prevent the deepfakes and
make the society ready to encounter the deepfakes. (Maras & Alexandrou, 2019)
According to an article published by, stated that it is not all the time the political
deepfakes mislead the individuals, but it will definitely implant a thought of uncertainty
which will impact the trust levels on the news media. It may be still at initial state, however
over a period, it will definitely affect the online civic culture and is potential enough to
produce provoking patterns among the viewers. Damage is not only measured in
monetary terms but also psychological terms when it comes to deepfake misuse. They
have conducted experiments to highlight the significance of the role played by the
deepfakes in eroding trust in social discourse while contributing to the misinformation that
28
is found online. They have taken large group of people from United Kingdom’s population
to gather the data required to analyze the people’s assessment of deepfakes and found
that the deepfakes makes people feel much more uncertainty than being misled and this
uncertainty will in turn induce indeterminably and cynical behavior among the people
and may even create trust issues on social media content and further create challenges
in retaining the online civic ethics in democratic communities (Vaccari & Chadwick, 2020).
According to Nicholas O’Donnell, in one of his published articles, deepfakes,
almost by definition are false and misleading in nature and the court has marked them to
be unprotected. He has classified the deepfakes into two basic categories such as
•
Deepfake pornography is the first category, which has an adverse effect
irrespective of the number of viewers to the deepfake videos.
•
The second category includes those deepfake media which requires certain level
disclosure and distribution to have a destructive impact.
The threats which are caused by the deepfake content have mainly divided into three
categories such as Elections manipulation, economic interference, and public safety
issues. According to Nicholas, the legal procedures which are available currently are
inadequate for confronting the threats associating with deepfake due to the presence of
Section 230 of Communication Decency Act that supports the social media corporations
in the cases where the accountability for the deepfakes that are spread virally on their
platforms and the recent amendments do not provide much effect as the social media is
immune from being liable to the information that is being posted and shared on their
platforms. Even supposing that there is an amendment in place for the section 230, any
rules applied to the deepfakes must be in compliance with the first amendment which
29
would be a weak regulation as the deepfakes come under a category of video editing
technology and also considered as a right to expression.
The above note is pertaining to the second category of the deepfakes and it also
argues that the Section 230 of Communication Decency Act should be amended in such
a way that it should consider the social media and online platforms as the disseminators
of the deepfake content which is displayed on their applications which makes the federal
regulations capable of penalizing the social media companies for the violation and can
demand them to pay the fines for disseminating the false information like deepfakes. In
most of the cases the fines issuance in large figures would be much more efficient than
civil and criminal system of liability as this will force the organizations to remove or get rid
of the offending deepfake media as soon as they are posted, and this kind of regulations
can withstand the constitutional challenge.
However, the introduction of the Communications Decency Act (CDA) has
provided the social media platforms with immunity against penalties and raised the issues
at policy level and created constitutional impediments that any deepfake content will face
and because of this, there is a need for the amendment to the Section 230 so that the
social media platforms will be held liable (O’Donnell, 2021). Assistant Professor and a
Deepfake pioneer, Hao Li says “This is developing more rapidly than I thought. Soon, it’s
going to get to the point where there is no way that we can actually detect [deepfakes]
anymore, so we have to look at other types of solutions.”
Talking about the rapid
development of deepfake and the need to develop the combatting methods for the misuse
of the deepfakes.
30
According to a study by Mika Westerlund (Westerlund, 2019), in which a review
and analysis on 84 recent public news on deepfakes was done in order to assimilate the
concept of deepfakes, who creates the deepfakes and its benefits and threats to the
humankind and the current examples of the deepfakes and mitigative measures to
contend them. The study has enabled them to know that the deepfakes are videos that
are digitally maneuvered to look hyper-realistic so that they portray people in a way that
they say and do the things that never occurred.
The study also found that, deepfakes are created using the GANs which can
generate new content based on the existing data and these deepfakes of real people
often tends to go viral and disseminates quickly on social media platforms and there by
acts as an effective tool for the disinformation. The study also discovered the various
contributions to the knowledgeable literature on the deepfakes which argues that
deepfakes are usually promoted because of the dependency of citizens on the
commercial media platforms, any deepfakes produced on heated conversations in the
political context, usually false in nature can be easily disseminated online and the
deepfakes gets its significance from the ability to utilize the advanced technologies like
Artificial intelligence to produce hyper-realistic videos.
The study supports these arguments by indicating that commercial platforms,
comprising both news media and social media platforms, are hotbeds for the production
of deepfakes. These deepfakes are not solely based on heated political arguments but
also on the broader social media context, leveraging the vast amount of online data
available. This situation can erode trust in data shared on social networks and even lead
people to doubt their prior beliefs.
31
The rise in the number of occurrences of fake news business models, which
generate a significant amount of web traffic through advertisements, is well-documented
in the study. This is supported by research analyzing news articles from journalists who
sometimes rely on unethical techniques like clickbait.
Furthermore, the study identifies several factors associated with deepfakes. These
include overly sensitive areas like governments, political extremists, lawbreakers, and
vindictive individuals who create fake media to provoke online paid and unpaid trolls.
Automated bots play a role in spreading this information on social media platforms. These
actors are primarily motivated by intentions to harm others in several ways and to
influence people to change their opinions on specific topics, leading to confusion in the
public. Their ultimate goals may include financial gain or altering opinions about
organizations, often for amusement or plain fun, which can impact individuals’ lives.
Despite its valuable findings, the study has limitations. It analyzed a limited number
of articles, specifically eighty-four online articles, to explore the concept of deepfakes.
This number is relatively small given the ever-growing development of technology. More
comprehensive insights might have been gained if a few more articles had been included
in the analysis.
Secondly, the sources taken into consideration in the study included only the public
sources like the online news sites for the article review and if other types data would have
been included in the study like the online community discussions on deepfakes,
considering the data gathered from the interviews given by the GAN developers and the
artists in deepfake creation field where few of the artists are recognized as not only
developers of the deepfake technology but also anti-deepfake technology developers who
32
can provide further insights on the policies to be followed to overcome the issues of the
deepfake and to mitigate the impacts of the deepfake in the lives of people. It also lacked
the inclusion of opinions and views as the study did not include the commentary sections
from the public news articles on the deepfake technologies so that the study could have
analyzed the ideas of the readers. The study also could have formed the insights on how
the deepfakes are perceived among the large audience so that there can be proper
information available to work on the methods to combat the impacts of deepfake and can
emphasize on the actual problem areas instead of beating around the bush. These
impediments have paved the way for the new research in the field of deepfake as there
is much more to be analyzed to overcome the issues of the deepfake.
Another study, Social Impact of Deepfakes by Dr. Hancock and Dr. Beilenson
discusses the way researches are conducted in the field of deepfakes as it is a recent
invention and still have scope in research despite the popularity of the deepfake media
and technology like the Faceapp application and Zao App on social media and online
platforms (Doffman, 2019). However, at the time of their research, only few studies have
been taken into consideration which analyzed the aspects of psychological, societal and
consequences of the policies that were in place at that time in a world where the media
can be easily shared across the world instantly whose legitimacy is unknown in most of
the cases and which are imperceptible to be real or fake (Hancock & Bailenson, 2021).
The need of their study was based on the face that there has been aplenty of
research done on the methods to detect the deepfake but there aren’t much research and
studies held on the impact of deepfakes on the society’s psychology and hence they come
up with an idea to study and examine the social impact of the deepfakes and the possible
33
impacts of deepfakes on the people. The study was mainly inspired by the Seitz’s and his
colleague’s presentation on the sensational deepfake video of then President Obama,
where a high-quality lip synching was achieved using deepfake where the mouth
movements from the younger Obama was used to create a deepfake video of Obama
twice his age which was perfectly drafted deepfake video with the help of machine
learning techniques. (Suwajanakorn et al., 2017). There were two main things that were
identified in the Seirz presentation which were discussed in this study which are as
follows:
•
The aspect was the fact that the algorithms used to generate deepfakes are much
easier than the algorithms used to detect them as per the basic nature of the GANs.
The deepfakes have been noticeably migrated from the computer programs in
science Laboratories to mere mobile apps which made it easy to generate a
deepfake content by a commoner who do not need much expertise in the field of
deepfake creation. Thus, the creation of deepfake has much advantage over the
detection of deepfakes.
•
The second aspect was the social and psychological impacts of deepfake and are
there any proper studies made on this aspect of deepfakes.
These were the main aspects that led to the initiation of this study. The study discussed
the concerns in the field of research on the deepfake technology such as:
1. Scarcity of empirical research: At the time of this study being conducted, there
were ample studies made on the creation of methods to detect the deepfakes.
However, there were not many studies and research made on the social impact of
the deepfakes where there exists rapid dissemination of deepfakes in the digital
34
world which are usually indistinguishable from the real videos. Although there were
various studies held on social prejudices and memory attainment from the altered
still photos, the psychological consequences of watching the AI crafted videos
have remained substantially remain unstudied. Peculiarly, the virtual reality has
been a great starting point for studying the social impact of the deepfakes. In virtual
reality, there is a feature where doppelgangers in the form of 3D models of a given
individua from the 2D images and once the doppelgangers are built, its much
simpler to generate the animations of this 3D models and rendered as the 2d
videos which looks indistinguishable with the real videos. Watching VR videos can
cause false memories which makes the individuals believe that they have
performed the deepfake activity and can even influence their preferences on the
brands the doppelganger uses in the VR (Segovia & Bailenson, 2009).
2. Few observations from the deception research: The Fundamental nature of the
deepfakes is deception which involves intentional, deliberate misleading of an
individual. The Study suggests that according to the literature on deception, people
tend to believe false evidence and are not usually good at detecting deception
while reading the messages. Also, studies have shown that the level of deception
detection is almost the same as in the case of messages and audio messages or
video clips.
The study found that it is usually surprising that accuracy of detecting
deception is much low as the people tend to believe what others convey without
much effort. Also, it is found that deception happens much easier when visual
media like videos as humans rely much on visual aids while perceiving information.
35
Although people easily believe the deception in the form of videos, knowing the
fact that deepfake videos exist might change their perception further as it may
create a doubt on believing the video contents to be real and it may certainly
interfere with the ability to acquire knowledge in the world of deepfakes. This may
in turn undermine the role of journalism and other media in the current digital world.
3. Aftermaths of the Deepfakes: According to an empirical study by Vaccari and
Chadwick (Vaccari & Chadwick, 2020), it is found that there is a rise in the sense
of uncertainty when the individuals have to trust the news as they are aware of the
existence of deepfakes similar as in the case of spam emails where the individual
easily ignores the email after becoming aware of such scams.
Hence, viewers are found to be developing resilience to the new forms of
deception like deepfakes. One of the most important concerns of using deepfakes
is the nonconsensual victim being portrayed in the videos or images usually as in
the case of pornography alterations in which they have never engaged. The impact
of such incidents can be devastating on the lives of the victims as the main
motivation behind their creation is they will be mainly used to humiliate, extort, or
harass the victims.
4. Future scope for the research: This study urges the researchers to conduct studies
on the social impacts caused by deepfakes as there is a remarkable development
in this field and can produce some interesting insights from the research as the
current study although being exploratory, but it is just preliminary. The study also
indicates that there is a need for the attention on frontier which is usage of the AI
powered filters that allows modification of the videos in real time usually the act of
36
smiling which shown positive effects on the viewers and stopping them to detect
the deepfake effects most of the time. The deepfakes, despite being able to
undermine the trust in the media and falsely manipulate the attitudes of the society,
it also turned into a more common place to use deepfakes in communication
context in daily activities and the study makes it clear that there is need for
conducting much more empirical research on the social and psychological impacts
of the deepfake as it is evolving like never before. (Hancock & Bailenson, 2021)
Summary
This chapter meticulously presents the context of the topic through background
delineation. It accentuates issues related to the mishandling and abuse of deepfake
technology, highlighting their severity. The chapter includes a detailed literature review
on existing legislation on deepfakes across the globe and underscores the need for
awareness on these issues. Furthermore, it uses statistics to depict the extent of
detrimental incidents that have occurred in the past. Additionally, the chapter describes
literature related to existing methodologies in the current topic.
37
Chapter III: Methodology
Introduction
The design of the study for the current topic is elaborated in a detailed
manner in this chapter including the methods followed for data collection. I have executed
the designated set of activities outlined in the Design Study to. Need to add content in
introduction.
Design of the Study
•
Research on the emergence of deepfake technology using online research platforms
like Google Scholar, SCSU Library and Research Gate.
•
Investigate the recent works to get the latest advancements and trends in deepfake
technology usage and its abuse.
•
Analyze the data and document in an organized way to present it to the readers in a
comprehensive way.
•
Analyze the obtained data and present the results in terms of the level of awareness
and suggest a technique or procedure to hinder the losses that can be caused by the
misuse of deepfake technology among the society. Need to modify this as per
agenda on how you are presenting the data.
•
Data Collection
The process of data collection comprises of two distinct phases: the first involves
a comprehensive literature review conducted on online research platforms, including
Google Scholar, SCSU Library, ResearchGate, and various miscellaneous articles from
reputable websites. – Need to add more content
38
•
Data Analysis
Need to add more content
•
Data Presentation:
Need to add more content
Tools Used: Need to add more content
Summary
This chapter provides a comprehensive summary of the research methodology
and study design, including Need to add more content. It outlines the data collection
procedures. The chapter also gives an overview of the data collection process to measure
the level of awareness on deepfake technologies. Furthermore, it delineates data analysis
procedures, focusing on identifying and presenting the collected data. Lastly, the tools
used in this research are briefly described.
39
Chapter IV: Results
Introduction
In this chapter, I will summarize the methodology employed to study the Privacy
and Security Implications of Deepfake Technology dissemination. This methodology
involves Need to add more content
Additionally, I will verify whether the research questions outlined during the
methodology phase have been adequately addressed in the study.
The chapter will present Need to add more content.
ADD tables, diagrams on current trends on awareness levels on Deepfakes.
Results and Analysis:
Need to add more content for RESULTS- Modify according to your data
Add more content to the literature review if needed to add results on the
study made on Implications of Deepfake Technology on Individual Privacy and
Security
The study highlighted the potential impact of deepfakes and the scenarios where
they could have the most significant consequences. Range of potential scenarios,
including political campaigns, celebrity scandals, and financial fraud, where the use of
deepfakes could have serious implications have been identified. The study indicated
pressing concern about the potential impact of deepfakes on individual privacy and
security, particularly in the context of intimate images or videos.
The study underscores the need for a multifaceted approach to address deepfake
risks as described below.
Preventive Measures and Regulations: Add content
40
Regulatory Frameworks: Add content
Education and Awareness: Add content
Public Awareness: Add content
Overall, addressing deepfake risks requires a comprehensive approach,
encompassing
prevention,
regulation,
education,
and
public
awareness.
This
multifaceted strategy can essentially mitigate the potential harms of this emerging
technology. Additionally, ethical, and legal considerations related to deepfake creation
and dissemination, including issues of privacy, consent, and intellectual property, should
be addressed.
Impact of Deepfakes on Democracy and Public Trust among the Society
Another key aspect to consider is the potential impact of deepfakes on democracy
and public trust. Deepfakes have the potential to undermine public trust in media and
information sources, which could have significant implications for democratic societies
(Chesney & Citron, 2019). As such, it is important to consider how deepfakes might be
used to manipulate public opinion and how to develop effective strategies to combat such
activities.
Finally, it is important to consider the broader context of technological change and
innovation in relation to deepfakes. As deepfake technology continues to evolve, it is likely
that new and more sophisticated forms of manipulation will emerge, making it increasingly
difficult to detect and combat the use of deepfakes. As such, it is important to continue to
invest in research and development to stay ahead of the curve and to ensure that
societies are equipped to address emerging challenges associated with deepfakes and
other emerging technologies.
41
Summary
After conducting the research throughout this study, the main objectives of this study
that are defined during the methodology have been achieved and addressed below.
1. Research on the emergence of Deepfake technology using online research
platforms like Google Scholar, SCSU Library and Research Gate.
The above-mentioned objectives of the study have been achieved by
conducting extensive research on the emergence of Deepfake Technology, its
usage and potential implications on society in terms of privacy and security issues
that are caused by the dissemination of deepfakes, by utilizing online research
platforms like Google Scholar, SCSU Library, and Research Gate various steps
has been taken.
2. Investigate the recent works to get the latest advancements and trends in deepfake
technology usage and its abuse.
Various relevant articles and publications were identified by searching on
these platforms with the relevant keywords like “deepfake technology”, “fake
videos” and “digital manipulation” etc. Furthermore, to refine the data for the study,
a date range and publication type filter is applied during the search for the relevant
information.
3. Analyze the data and document in an organized way to present it to the readers in
a comprehensive way.
A thorough study has been done on the abstracts and full texts of the
selected articles on the deepfake emergence, its usage and potential implications
are reviewed to determine their relevance and potential to contribute to the study.
42
The sources that were deemed relevant were included in the literature review
section of the study. In addition to the research platforms like Google Scholar,
SCSU Library and Research Gate that were used to gather information for this
study, other sources were used to conduct research on the emergence of deepfake
technology including news articles, government reports and other media resources
that provided the information on the major developments and use of deepfake
technology, related legislature and current mitigative measures to combat misuse
of deepfakes. The combination of various sources allowed for a comprehensive
analysis of the emergence of deepfake technology and its potential implications
for society and measures to be taken in case of misuse of the deepfakes.
4. Analyze the obtained data and suggest a technique or procedure to hinder the
losses that can be caused by the misuse of deepfake technology among the
society.
Need to address 5th point, how I have achieved this objective from
design of study
43
Chapter V: Conclusion and Recommendations
Conclusion
This paper presents and overview of deepfakes and discusses its societal impact
on security and privacy, highlighting its challenge in discerning real media from fake
media. Furthermore, the emergence of deepfakes poses significant societal and business
challenges, potentially harming individuals, and eroding media trust. This paper has also
highlighted the necessity for regulations, detection, mitigation, and prevention measures
against deepfake misuse. The opportunities available to cybersecurity and AI to combat
fake news and media have been discussed and it emphasized the need for addressing
and increasing awareness on potential dangers of deepfakes among the policymakers
and public. The obtained data on Deepfake awareness among have been analyzed to
classify the audiences into different groups of varying awareness levels on deepfake
technology. Suggestions have been made to address the lack of awareness among the
target audiences by providing possible and the need for the effective ways to overcome
the lack of awareness of this novel technology among the society and a future scope has
been defined to lay a path for the future possible research on the implications of
deepfakes on the security and privacy of the individuals in a society.
Although this project may not solve all the issues of deepfake technology, it aims
to spread awareness and equip the public with knowledge to defend themselves against
it. The long-term solution involves critical thinking and research, aided by technology. The
project has aimed to contribute to this solution by exploring the power of design to reveal
the truth and create awareness.
44
Recommendations
As the target audience include people from diverse backgrounds, the type of training
that can be given will depend on the kind of target audiences and their level of knowledge
and skills. Upon thorough examination of potential target audience categories, the training
pertaining to deepfake awareness can be methodically classified into five primary
categories as described below:
1.
General awareness training: This type of training should be given for individuals
with limited knowledge about deepfakes is general awareness training. This
type of training would provide an overview of deepfakes, including their creation
process and the possible consequences they could have on both individuals
and society.
2.
Detection training: For individuals who are responsible for verifying or
authenticating media content, such as journalists or social media moderators,
this training could be beneficial. It could cover topics such as how to detect
deepfakes using various techniques, tools, and technologies.
3.
Prevention training: Individuals who are involved in creating or sharing media
content, such as social media users or content creators, could benefit from
prevention training. This training would provide guidance on how to prevent the
creation or spread of deepfakes by using secure authentication methods or
watermarking technologies.
4.
Legal training: Legal professionals such as law enforcement officials and
policymakers, as well as individuals in the legal field, could benefit from legal
training on deepfakes. It could cover topics such as the legal implications of
45
deepfakes, including issues related to privacy, intellectual property, and
defamation.
5.
Ethical training: Individuals who produce or consume media content,
policymakers, and educators could benefit from ethical training on deepfakes.
This training would focus on ethical considerations related to deepfakes,
including topics such as consent, manipulation, and bias.
Future Work
Based on the findings of the study on the Implications of Dissemination by
Deepfake Technology on Privacy and Security of the individuals in a Society using
deepfake awareness study, some recommendations for future work and using the
study’s results are:
•
Conduct a larger-scale survey: Currently there are very smaller number of studies
on awareness levels on Deepfakes, so future work could involve promoting
conducting extensive surveys to obtain a broader range of perspectives and a
more representative sample.
•
Investigate the effectiveness of deepfake detection tools: As deepfake technology
evolves, there is a need to develop effective detection tools to mitigate the risks
associated with their use. Future work could investigate the effectiveness of
various deepfake detection tools and how they can be improved.
•
Explore the ethical and legal implications of deepfake technology: Deepfake
technology raises a range of ethical and legal issues, such as privacy violations,
identity theft, and defamation. Future work could explore these issues in more
detail and identify ways to address them.
46
•
Develop deepfake awareness campaigns: Given that many individuals are
unaware of the existence and potential harms of deepfake technology, future work
could focus on developing public awareness campaigns to educate people about
the risks associated with deepfakes and how to protect themselves.
•
Evaluate the effectiveness of deepfake awareness interventions: Future work
could also evaluate the effectiveness of deepfake awareness interventions, such
as educational campaigns or workshops, in increasing people’s awareness and
ability to protect themselves against deepfakes.
•
Identify and mitigate the social and political impacts of deepfakes: The
dissemination of deepfakes can have serious social and political consequences,
such as misinformation, manipulation, and interference in elections. Future work
could focus on identifying these impacts and developing strategies to mitigate their
effects.
Overall, the study’s results can be used to inform the policymakers, educators, and
the public about the potential harms of deepfake technology and the need for greater
awareness and mitigation strategies for the security and privacy issues caused by the
dissemination of the deepfakes in the current AI driven world.
47
References
Antigone, D. (2021, December 2). Strengthening Our Efforts Against the Spread of Non-Consensual
Intimate Images. Retrieved from about.fb.com:
https://about.fb.com/news/2021/12/strengthening-efforts-against-spread-of-non-consensualintimate-images/
Artz, K. (2019, October 11). Texas Outlaws “Deepfakes”—but the Legal System May Not Be Able to Stop
Them | Texas Lawyer. Retrieved from www.law.com:
https://www.law.com/texaslawyer/2019/10/11/texas-outlaws-deepfakes-but-the-legal-s
Bregler, C., Covell, M., & Slaney, M. (1997). Video rewrite: Driving visual speech with audio. Proceedings
of the 24th annual conference on Computer graphics and interactive techniques. (pp. 353-360).
ACM Press/Addison-Wesley Publishing Co.;.
Briscoe, S. (2023, July 24). U.S. Laws Address Deepfakes. Retrieved from Asisonline.org:
https://www.asisonline.org/security-management-magazine/latest-news/today-insecurity/2021/january/U-S-Laws-Address-Deepfakes/
BURCHARD, H. V. (2018, May 21). Politico. Retrieved from https://www.politico.eu/:
https://www.politico.eu/article/spa-donald-trump-belgium-paris-climate-agreement-belgiansocialist-party-circulates-deep-fake-trump-video/
Chesney, B., & Citron, D. (2019). Deep Fakes: A Looming Challenge for Privacy, Democracy, and National
Security. CALIFORNIA LAW REVIEW, 1779-1783.
Chesney, R., & Citron, D. K. (2018, July 14). Deep Fakes: A Looming Challenge for Privacy, Democracy,
and National Security. SSRN, p. 1785.
Çolak, B. (2021, January 19). Legal Issues of Deepfakes. Retrieved from internetjustsociety.org:
https://www.internetjustsociety.org/legal-issues-ofdeepfakes#:~:text=However%2C%20deepfakes%20have%20some%20drawbacks,rights%20and
%20intellectual%20property%20rights.
CVISIONLAB. (n.d.). Deepfake (Generative adversarial network). Retrieved from www.cvisionlab.com:
https://www.cvisionlab.com/cases/deepfake-gan/
Doffman, Z. (2019, January 27). Chinese deepfake app ZAO goes viral privacy of millions ‘at risk’.
Retrieved from https://www.forbes.com/:
https://www.forbes.com/sites/zakdoffman/2019/09/02/chinese-best-ever-deepfake-app-zaosparks-huge-faceapp-like-privacy-storm/?sh=644d2b2e8470
48
Fagni, T., Falchi, F., Gambini, M., Martella, A., & Tesconi, M. (2021). TweepFake: About detecting
deepfake tweets. Plos.org. Retrieved 2023, from
https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0251415
Hancock, J. T., & Bailenson, J. N. (2021, March). The Social Impact of Deepfakes. Cyberpsychology,
Behavior, and Social Networking, pp. 149-152.
Hughes, S., Fried, O., Ferguson, M., & Hughes, C. (2021). Deepfaked online content is highly effective in
manipulating people’s attitudes and intentions. TBD, 2-4.
Jia, Y., Zhang, Y., Weiss, R. J., Wang, Q., Shen, J., Ren, F., . . . Wu, Y. (2018). Transfer learning from
speaker verification to multispeaker text-to-speech synthesis. Proceedings of the 32nd
International Conference on Neural Information Processing Systems (pp. 4480-4490). Curran
Associates Inc.
Lovells, H. (2020). Deepfakes: An EU and U.S. perspective. Retrieved from GMCQ__Spring_2020_Deepfakes: https://f.datasrvr.com/fr1/320/16758/1207330_-_GMCQ__Spring_2020_Deepfakes.pdf
Maras, M.-H., & Alexandrou, A. (2019). Determining authenticity of video evidence in the age of artificial
intelligence in the wake of Deepfake videos. International Journal of Evidence & Proof, 255-262.
O’Donnell, N. (2021). HAVE WE NO DECENCY? SECTION 230 AND THE LIABILITY OF SOCIAL MEDIA
COMPANIES FOR DEEPFAKE VIDEOS. UNIVERSITY OF ILLINOIS LAW REVIEW.
Petkauskas, V. (2021). Report: number of expert-crafted video deepfakes double every six months.
CyberNews.
Reid, S. (2021, January). https://www.scholarship.law.upenn.edu/. Retrieved from The Deepfake
Dilemma: Reconciling Privacy and First Amendment Protections:
https://scholarship.law.upenn.edu/jcl/vol23/iss1/5/
Segovia, K. Y., & Bailenson, J. N. (2009). Virtually true: children’s acquisition of false memories in virtual
reality. Media Psychology, 371-393. Retrieved from https://stanfordvr.com/mm/2009/segoviavirtually-true.pdf
Stupp, C. (2019, August 30). Fraudsters Used AI to Mimic CEO’s Voice in Unusual Cybercrime Case.
Retrieved from https://www.wsj.com/: https://www.wsj.com/articles/fraudsters-use-ai-tomimic-ceos-voice-in-unusual-cybercrime-case-11567157402
Suwajanakorn, S., Seitz, S. M., & Shlizerman, I. K. (2017). Synthesizing Obama: Learning lip sync from
audio. ACM Transactions on Graphics. Retrieved from
https://grail.cs.washington.edu/projects/AudioToObama/siggraph17_obama.pdf
Tremaine, D. W. (2019). Two New California Laws Tackle Deepfake Videos in Politics and Porn. Retrieved
from www.dwt.com: https://www.dwt.com/insights/2019/10/california-deepfakes-law
49
Vaccari, C., & Chadwick, A. (2020). Deepfakes and Disinformation: Exploring the Impact of Synthetic
Political Video on Deception, Uncertainty and Trust in News. Social Media + Society, 1-13.
Virginia Legislative Information System. (2019). Bill Tracking – 2019 session > Legislation . Retrieved from
https://lis.virginia.gov/: https://lis.virginia.gov/cgibin/legp604.exe?191+ful+CHAP0490&191+ful+CHAP0490
Westerlund, M. (2019). The Emergence of Deepfake Technology: A Review. Technology Innovation
Management Review, 39-47.
WIPO. (2019, December 2019). Draft Issues Paper on Intellectual Property Policy and Artificial
Intelligence. United States of America.
Zeman, E. (2021). Insurers Face Evolving Cyberrisk From Costly Hacks, Deepfake Attacks and
Sophisticated Ransomware. Best’s Review, 48-52.