Survey Paper on Content Moderation

With the growth and accessibility of the internet to everyone, recognizing pornographic images is of great significance for protecting children’s physical and mental health. It’s common for children to surf the internet and they are just one click away from getting access to pornographic images. It is the desire of all parents to protect their children from online pornography, cyberbullying, and cyber predators. Several existing methods analyze a limited amount of information from a child’s interaction with the respective online section. Some restrict access to sites based on blacklists known as banned URLs, others attempt to scan and analyze media content being exchanged between two parties. However, new URLs can be used to bypass blacklists, and images, videos, and text appear to be safe individually but should be evaluated together. Due to the vast size of social media-generated content, these approaches are insufficiently accurate. Furthermore, the approaches are not discriminative enough on a variety of image properties. The Colour, shadow, and frequency features of the images can vary, even the context is the same according to lumination features. The problem can be solved more accurately with deep learning techniques. Notably, the specific type of deep learning architecture called convolutional neural network is suitable for the problem space. Having a potential solution becomes even more important than detecting and recommending NSFW (Not Safe For Work) content that may be present to the user.