Cyber-Flashing Identification Method of Analysing Human Skin on Social Media Platforms

The use of social media, such as Facebook, Twitter, WhatsApp, WeChat, and others have grown exponentially over time with the growth of the Internet and has become the most influential networking platforms in the 21st century. They are well-known for their low-cost, quick, and effective communication. However, the enhancement of social connectivity often creates negative impacts on society that contributes to online abuse, harassment, cyberbullying, cyber flashing, cybercrime, and online trolling. In this paper, we are focusing on cyber-flashing, the unsolicited sending of nude or sexual images, which is a common practice but often met with negative reactions from receivers. During and after COVID lockdown, cyber-flashing has been a serious threat on WhatsApp recently. The majority of study, meanwhile, has focused on identifying instances of cyberbullying rather than the effects of cyber-flashing. Cyber-flashing is the act of sending an inappropriate photo or video to another person, as well as exposing oneself to another via live video. An essential step in the detection of cyber-flashing is evaluating the level of risk in terms of abusiveness in the provided picture, or video. Therefore, the goal of this research is to create and refine a method that can effectively identify abusive cyber-flashing content and prevent inappropriate content sharing within the social media platforms or public Wi-Fi network. As a result, we have proposed a unique technique, where we employ a multi- stage model for identifying cyber-flashing photos and the proposed YOLOv3-YCbCr performs best in terms of metrics, with 98.14% accuracy, 95.58% specificity, and 97.27% sensitivity. First, the YOLOv3 technique is used to detect human presence in the image. Then, the YCbCr model is used for the skin detection and analysis. In the context of intelligent systems, skin detection is a difficult topic that has drawn a lot of attention from specialists and experts from the research community; the lack of uniform testing methodologies and standard benchmarks have made it difficult to compare different approaches fairly. Based on our studies, YOLOv3-YCbCr performs far better for skin segmentation than any prior hand-crafted approach method.