Trust Evaluation with Deep Learning in Online Social Networks: A State-of-the-Art Review

In the realm of online social networks (OSNs), it has become increasingly crucial to analyze user behavior, establish trustworthy relationships to mitigate social risks, enhance security, and safeguard privacy. Trust evaluation is widely acknowledged as an effective approach for detecting internal attacks and identifying compromised nodes, and deep learning technology can significantly enhance its performance. However, there remains a notable gap for a review paper focused on trust evaluation utilizing deep learning techniques within OSNs. Therefore, conducting a state-of-the-art review on this subject has become imperative. We analyze and compare some recent related research, summarizing prevalent challenges and open issues while proposing optimization strategies to address them. For instance, graph-based neural networks methods often grapple with exponentially increasing computational complexity as network size expands, and imbalanced datasets typically lead to reduced model accuracy and generalization. Lastly, it presents several promising avenues for future research in the field.