Enhanced Soft Computing Techniques for Fake News Detection

The rapid spread of fake news on social media poses a signif- icant cybersecurity threat, necessitating early detection and effective solutions. This research proposes an energy-efficient and explainable multimodal fake news detection (FND) frame- work that integrates textual, visual, and user-behavioral cues for enhanced reliability. Two hybrid architectures, LSTM + ANFIS and LLM + AutoML, were developed and evaluated to balance accuracy, training time, and interpretability. Re- sults across benchmark datasets (Twitter, PolitiFact, and Buz- zfeed) show that the LSTM + ANFIS model achieves up to 96.1% accuracy while reducing training time by over 50%using F-Transform, offering superior energy efficiency and trans- parency compared to existing approaches. To enable early detection and mitigation, the study further incorporates a fuzzy rule–based SEI epidemiological model, which simulates user behavior and information propagation to identify potential spreaders before large-scale diffusion. The integration of fuzzy inference and explainable reasoning provides interpretable, human-understandable insights, en- hancing trust in automated systems. Overall, the proposed framework contributes to a sustainable, interpretable, and proactive solution for early fake news detection and cyberse- curity resilience in multimedia-driven digital environments.