Introduction to Article Review 1

Article Review #1: “Testing human ability to detect ‘deepfake’ images of human faces”
Zephaniah Hall
School of Cybersecurity, Old Dominion University
CYSE 201S: Cybersecurity and the Social Sciences
Instructor name: Professor Diwikar Yalpi
February 26th 2026
To sum up the article, certain public and private websites are posting deepfake using the
StyleGAN2 generation method to increase the cause of concern for deepfake images.

Increasing the urgency of this threat relates to many different images, and other types of
information that could be displayed online. Many different types of information could be
changed, altered, or opinions swayed with deepfake images. Information travels fast, and if the knowledge of the deepfake images aren’t available then it could cause a lot of widespread
damage in a lot of ways. In fact being that, the images appeal to one of the senses which is the
sight, there is only so much information we can receive that isn’t empirically and the deepfake
attacks the sight, and even certain audios, so the knowledge helps us to learn to use more than
our sight, and our mind to discern the images.

The main questions addressed in this article is: Can people differentiate the difference
between deepfake (AI generated) and authentic images of people? And will advice about how to detect a deepfake improve performances? The hypothesis is that either machine learning systems can detect pattern recognitions in deepfakes or blockchain technologies to create white lists of non-deepfake instances. The experiment that was conducted was a group of participants (280) with the information that they were to determine the differences between real and fake images, eventually they were randomly allowed into one of four study conditions with a baseline control condition, and 3 special experiment conditions. The first was a familiarization intervention and participants were shown 20 examples of deepfake images and were knowledged that there were.

The same 10 images, weren’t used for the experimental task but for each experiment with
randomized orders. In the second and third experiments, participants were shown a list of 10
telltale features that the deepfake images commonly shared and were used to distinguish from
non-deepfake pictures. For the second condition, the advice was shown before the experiment
started only, and for the third, an additional reminder of the names of each tell-tale feature was displayed beneath each image throughout the experiment. For the Dependent Variables, different tables were shown with information like the mean number of times participants correctly identified as real or fake (participants were correct 60% of the time), and a more detailed breakdown showing accuracy for real and deepfake images separately, with the data for the reminders, one time advised, and control groups. Table 4 showed the number of deepfake stimuli observed across conditions, along with the mean number of deepfake labels applied. Table 5 showed the participants’ mean confidence in their decisions, and information on the mean accuracy rates per image.

The type of research methods collected were mixed involving qualitative and quantitative
so a mixed collection, also with a good amount of diversity having people of all ages and uk, non uk, english, european and us participants. Based on the course concepts, I can see the basic concepts with experimental treatment and treatment manipulation. The manipulation could be done in the experiment to alter perceptions, not with data but the way the answers are given. The study focuses on normal individuals who could be susceptible to being deceived by deep fake and ai. It advances our understanding of cybersecurity and the social sciences by giving information of how to learn and have the knowledge of the false videos/images. Being aware of them can cause more precautions to be taken looking at medias. Even having more precautions can make us slow down and take our time analyzing media and allow us to realize how incorrect or not real things are.
Source:
Bray, S. D., Johnson, S. D., & Kleinberg, B. (2023). Testing human ability to detect ‘deepfake’
images of human faces. Journal of Cybersecurity, 9(1), tyad011.
https://doi.org/10.1093/cybsec/tyad011