The fast growth of digital communication has made it more difficult to arrest those who breach the law. Criminals are increasingly working together behind the scenes on social media platforms to spread false information, perform criminal crimes, and avoid being discovered. Traditional forensic procedures cannot extract valuable information from the overwhelming majority of data collected from social media platforms because it is unstructured. Hence, this paper proposes a Deep Learning-based Social Network Forensic Data Mining (DL-SNFDM) to automatically gather evidence of people’s interactions on graphs using analytics and deep neural networks. For the purpose of meaning interpretation, DL-SNFDM uses transformer-based language modeling, Convolutional Neural Networks (CNNs) for visual data sorting, and Social Graph Inference (SGI) to identify powerful nodes, hidden connections, and groups of people who behave in an aberrant manner. Experimental findings show that DL-SNFDM significantly reduces the time to acquire evidence by 32%, improves the accuracy of detecting signals of criminal intent by 19%, and makes it easier to discover connections between suspects at the network level. Moreover, the model would reveal communication hierarchies and social patterns that are not reflected in rule-based systems. Finally, the DL-SNFDM framework, which provides law enforcement with an intelligent, automated, and scalable forensic model, contributes to simplifying cutting-edge cyber-assisted criminal investigations.
F. Yan (Wed,) studied this question.