ABSTRACT
Networking is the swapping of data or views among people or professionals through various social networking platforms. Online social networks (OSNs) play a major role in everyone’s life.
Owing to the increase in the number of OSNs, people can communicate with people from all over the world. Public social networks, including Facebook, X/Twitter, LinkedIn, and Instagram, allow people to post their thoughts and apply for jobs.
Spam bots play an important role in these OSNs. They are nothing but computer programs designed to send spam over the Internet. Online social networks can provide negative consequences if people all over the world misuse them, and one of the major negative activities that happen in OSNs is cyberbullying.
Cyberbullying is a major threat that spreads among OSN users, particularly teenagers. Cyberbullying refers to the threat and harassment of OSN users.
Both spam bots and cyberbullying must be detected to protect OSN users from various attacks. The majority of existing studies are based on data communication and networking intrusion detection, utilizing different properties and employing unique and fusion classifiers.
Owing to dataset imbalance and poor data modeling, these techniques did not work. Because of this drawback, many researchers have developed decentralized models to detect spam bots and cyberbullying in OSNs.
In addition, some deep learning models, such as recurrent neural network, bidirectional long short-term memory, convolutional neural network, and long short-term memory, have been used to detect spam bots and cyberbullying.
This systematic review deals with the overall view and challenges of earlier and recent deep learning methods to detect spam bots and cyberbullying in OSNs for future research.
