Social media is one of the topmost communications in the present world and some discomforts are also hidden in this. Social media is one of the platforms for cyberbullying. Teenagers who are not under the guidance of their parents are the major victims of cyberstalking and cyber abuses. Forensic image processing enhances the digital image using different computer techniques. Deep fake uses the technology to superimpose more than one image, video, and audio with other images, videos, and audios with the well-known artificial intelligence and generative adversarial networks (GANs). This technology raises the curtain to create many fake images, fake videos, especially porn videos and fake audios of celebrities, especially actors and politicians. This chapter explains the impacts of deep fakes in social media, especially Facebook and Twitter. This chapter proposed a new framework called the extended-RACYBDD to detect the deep attacks.
TopIntroduction
In today’s technology-driven world, people use social media for various reasons, including conducting business, communicating with family and friends, and receiving the news. Many are addicted to conformity with this media, which can be problematic, especially among shared accounts. Whenever friends, family, individuals, and colleagues meet using social media, that choice directly affects their social media presence and how they identify within a particular platform. If an individual does not have social media, they are often pressured by friends, family, or colleagues propose to join social media because of all of its purported communicative advantages. With the ongoing improvement of GAN (Generative Adversarial Networks) and its variants, the images may experience the ill effects of being controlled by different GAN's to make deepfakes (Abhijith, 2019). Mishandling the deepfakes may bring unexpected results. So, decisive measures need to be taken to fight against deepfake to secure the images on social media. Deep learning technologies can be used to detect deepfakes (Sheng-Yu, 2020). Without a doubt, progressed GAN's will be created to deliver excellent the modified images with fewer antiquities. These serious GAN's will be applied for making deepfakes malevolently and present genuine difficulties for discovery since existing identifiers are not summed up to obscure GANs. Moreover, ongoing investigations have exhibited that Deep neural networks (DNN) based identifiers are defenseless to ill-disposed commotion assaults by adding unsavory imagery into the original face images. To address these issues, a novel technique named Extended Ripple and Cyberbullies Data Detection for Deep Fakes(E-RACYBDD) framework has been proposed.
The main contributions of this chapter are as follows:
- 1.
First, understand the infiltration of deepfakes and cyberbully detection in social media platforms, automatically identifying bullying persons by detecting deepfakes.
- 2.
Propose a novel Extended RACYBDD Framework to defend images on Social Media by giving confidentiality to the individual photos that are distributed on social media.
- 3.
Develop an Extended replica algorithm to not make the exact copy of the images from the social media when cyberbullies are trying to save or share others' profile personal photos from social media.
- 4.
Develop an Extended RACYBDD Algorithm to reveal the details of the cyberbullies when the cyberbullies are trying to upload the deep fake photos downloaded from social media.
- 5.
Conduct experiments on three real-world data sets from Facebook, Twitter, Instagram. Results show that the Extended RACYBDD model outperforms the state-of-the-art models and provides privacy and security to the uploaded images in social media.
The rest of the chapter is organized as follows: Section 2 focused on the emerging techniques in the creation of deepfakes. Section 3 describes the RACYBDD system architecture, while Section 4 introduces the Extended Replica algorithm. In Section 5, the Extended Cyber Bullies Data Disclose algorithm is discussed. Finally, Section 6 discusses the experimental results and their performance with areas of future research provided in the conclusion section.