Adaptive Defense Mechanisms for Rumor Detection on Social Media: A Contrastive Learning Approach

Abstract

The proliferation of rumors and misinformation on social media has emerged as a critical challenge, influencing public opinion, undermining decision-making processes, and threatening social cohesion. To address this issue, the present study proposes an adaptive defense mechanism for rumor detection based on a contrastive learning framework. The method is specifically designed to enhance both the accuracy and robustness of rumor detection models operating in dynamic and noisy online environments. By leveraging contrastive learning, the model captures and encodes spatial relationships among social media posts, enabling more precise differentiation between rumors and non-rumors. This enriched representation supports the development of more resilient models capable of maintaining high performance under adversarial conditions. The model’s effectiveness is assessed through comprehensive experiments that evaluate classification accuracy, resilience to adversarial attacks, memory usage, and computational efficiency. Results indicate that the contrastive learning-based model achieves a high classification accuracy of 92.7%, outperforming traditional machine learning techniques. Furthermore, the model maintains strong performance when exposed to adversarial inputs, demonstrating its robustness. Its memory-efficient architecture and low execution time make it a viable solution for real-time applications across large-scale social media platforms. The findings emphasize the potential of adaptive, contrastive learning-based approaches in mitigating the impact of online misinformation. This study contributes to the growing field of intelligent misinformation detection by offering a scalable and efficient solution, paving the way for future advancements in real-time, automated rumor detection technologies.