Developing a Deep Learning Model to Classify Malicious User on Social Networks

The use of social media for business and politics has resulted in an increase in negative behavior. Social media engagement, including personal, corporate, and political propaganda, has encouraged individuals to engage in actions that may not be in their best interests. Malicious users have been able to take advantage of hacked profiles to spread criticism with little consequence. This study aims to identify any issues by analyzing important aspects of social media data, uncovering complex connections to identify problematic behavior, such as fake or malicious social network accounts. The research employs Influence, Homophile, and Balance Theory within a social framework to enhance the accuracy of classifying dangerous users. The Jaccard coefficient is utilized to measure similarity, while graphical and linguistic cues are used to classify end-users in User-space. The framework is assessed using standard parameters such as the confusion matrix to determine its efficacy. For social atom anomaly detection, the friend connection identification framework is recommended.