Online social networks (OSNs) have gained a lot of popularity in the past few years. The capacity of OSNs to provide a way for users to interact with their peers is the driving force behind this phenomenon. Social media sites like Facebook, Twitter, and Instagram have recently become an inextricable part of our everyday lives. Information sharing raises many security and privacy problems, in cases where users upload personal material such as photographs, videos, and audio. An attacker can take advantage of shared data for malevolent purposes. In the case of children, the risks are much higher. This study examines numerous risks associated with OSN and also possible ways that can secure social network users to address these issues. The research will also prove that DL is a viable and scalable approach for OSN's state-of-the-art PPS by using a deep neural network for identifying potential threats. The framework proposed in the chapter helps identify features related to attacks on Online Social Networks, determine the relevant policies, scan networks, and create subnetwork anomaly nodes. It’s a continuous process to perform retrospective analysis and help improve models by adjusting parameters. Using Deep Learning Algorithms, attributes are classified and stored in a secure database. This repository is used in detecting attacks on online social networks and taking the most appropriate actions. Finally, it identifies several unresolved concerns and obstacles that now impede real-world implementation and suggests future paths for achieving trustworthiness in online social networks along this dimension.