The integration of social dynamics into the Internet of Things (IoT) networks, termed Social IoT (SIoT), presents a challenging task with regards to trust management due to the dynamic and socially influenced nature of the SIoT networks. Classical trust models struggle to adapt to the complex SIoT environments, leaving the possibility of malicious attacks. This paper proposes a framework for the SIoT ecosystem, taking advantage of blockchain technology and Neural Networks to enhance trustworthiness assessment to mitigate risks. The proposed framework leverages blockchain for secure data storage and transaction transparency to ensure the integrity of the information. Neural network algorithms like Recurrent Neural Networks (RNN) and Bidirectional Encoder Representations from Transformers (DistilBERT) are used to assess trust in real-time, taking into account evolving social interactions, leveraging the advantage provided by transfer learning. The simulation-based experiments are conducted to evaluate the efficiency of the proposed framework for detecting and mitigating malicious attacks in SIoT environments. Results demonstrate the robustness of the solution.