Abstract
Deepfakes are realistic-looking fake media generated by deep-learning algorithms that iterate through large datasets until they have learned how to solve the given problem (i.e., swap faces or objects in video and digital content). The massive generation of such content and modification technologies is rapidly affecting the quality of public discourse and the safeguarding of human rights. Deepfakes are being widely used as a malicious source of misinformation in court that seek to sway a court’s decision. Because digital evidence is critical to the outcome of many legal cases, detecting deepfake media is extremely important and in high demand in digital forensics. As such, it is important to identify and build a classifier that can accurately distinguish between authentic and disguised media, especially in facial-recognition systems as it can be used in identity protection too. In this work, we compare the most common, state-of-the-art face-detection classifiers such as Custom CNN, VGG19, and DenseNet-121 using an augmented real and fake face-detection dataset. Data augmentation is used to boost performance and reduce computational resources. Our preliminary results indicate that VGG19 has the best performance and highest accuracy of 95% when compared with other analyzed models.
Highlights
Our preliminary results indicate that VGG19 has the best performance and highest accuracy of 95% when compared with other analyzed models
Given that the most successful approaches to identifying and preventing deepfakes are deep-learning methods that rely on Convolutional Neural Network (CNN) as the backbone for a binary classification task [12], and a large 2D CNN model can prove to be better than EfficientNet model if deepfake classification is the only desired result [55], we have evaluated the most common backbone architecture of existing developed frameworks (CNN, VGG-19 and DenseNet) that are demonstrated to have the best performance on the task of deepfake image classification
The results of our work demonstrated that deep-learning architectures are reliable and accurate at distinguishing fake vs. real images; detection of the minimal inaccuracies and misclassifications remain a critical area of research
Summary
Giorgio GiacintoIn the last few years, cybercrime, which accounts for a 67% increase in the incidents of security breaches, has been one of the most challenging problems that national security systems have had to deal with worldwide [1]. Deepfakes (i.e., realistic-looking fake media that has been generated by deep-learning algorithms) are being widely used to swap faces or objects in video and digital content. This artificial intelligence-synthesized content can have a significant impact on the determination of legitimacy due to its wide variety of applications and formats that deepfakes present online (i.e., audio, image and video). Considering the quickness, ease of use, and impacts of social media, persuasive deepfakes can rapidly influence millions of people, destroy the lives of its victims and have a negative impact on society in general [1]. The generation of deepfake media can have a wide range of intentions and motivations, from revenge porn to political fake news. Deepfakes have been published to falsify satellite images with non-existent landscape features for malicious purposes [3]
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call