Vehicular social networks (VSNs) are a new type of communication where essential ideas from two distinct fields—mobile social networks (MSNs) and Vehicular Ad-hoc Networks (VANETs) are combined. In the virtual community of cars, passengers, and drivers on the roads, VSNs involve social interactions of vehicles with similar goals, interests, or movement patterns. Possessing a good ability for fraudulent content detection is useful in establishing improved network performance in VSN. In this paper, a hybrid Deep Neural Network-based model for fraudulent content detection for long-term VSN is proposed. Recurrent Neural Networks (RNNs) and Generative Adversarial Networks (GANs) are combined to capture two parts of semantics, using Graphical Neural Network (GNN) as the foundational architecture. Several tests are run on social network dataset for evaluation, and the findings showed that they could increase the detection effect by 8% to 15% over the baseline techniques.