Facial Expressions Analysis For Deep Fake and Genuine Video Recognition

Facial Expression analysis (FEA) is a process that involves recognition and understanding of human emotions based on facial cues. FEA has the potential to manipulate people’s appearance, statements, and opinion. This is important as FEA is becoming a more popular tool and when misused can lead to the spread of misinformation such as deepfake technology. Deepfake has emerged as a massive phenomenon and a result of FEA abuse. For example, many individuals find it challenging to distinguish between deepfake and authentic content. To address this issue, a research experiment was conducted to gain an insight into how people react towards deepfake and authentic contents. Respondents were shown videos and an analysis was conducted on participant’s facial expressions as well as assessing their knowledge of deepfake detection. 40 participants watched several videos, comprising both deep fake and authentic content and after the conclusion of each video, participants were asked to complete a survey. The survey was designed to test their confidence with the level of deepfake and authentic video identification, trust, security, and attitude towards them. Facial expressions were analyzed using Noldus FaceReader 7 to detect and classify 7 facial expressions (such as happy, sad, neutral, angry, surprised, disgusted, and other). The study findings indicate that FaceReader analysis discerns a statistically significant difference in emotional responses between real and deepfake videos, while participants reported a higher percentage of neutrality (70% vs. 62.5%) in real videos compared to deepfakes.