Research work · 2025
A Hybrid Intelligence Framework for EmotionAware Deepfake Detection and Misinformation Risk Reduction
2025 Third International Conference on Emerging Applications of Material Science and Technology (ICEAMST)
Abstract
The development of deepfakes has accelerated to increase the rates of misinformation, emotional control, and loss of digital trust. The current detection systems mainly aim at seeing or hearing something incongruent and overlook the emotional aspect that compounds psychological and social damage. To fill this gap, a Hybrid Intelligence Framework of Emotion-Aware Deepfake Detection and Misinformation Risk Reduction based on deep learning, affective computing, and explainable AI is proposed in this study. The framework is a combination of spectral spatial forensic, linguistic sentiment, and physiological emotion characteristics via a transformerdirected multimodal fusion network. The hybrid CNN BiLSTM framework structures spatial forgery localizations and emotional pathway dynamics, which are optimized through adaptive gradient control and attention-guided calibration. The experimental performance on multimodal deepfake datasets shows that the system detects correctly 96.4% with 0.97 AUC and a false emotional trigger rate is reduced by 21 percent over the state-of-the-art systems, which were correspondingly statistically significant ($\mathbf{p}{\lt}0.05$). The proposed model increases interpretability, stronghold, and ethical online management through the detection of emotionally manipulative synthetic media and the measurement of the risk of misinformation, which provides a starting point to acquire emotion-stable and reliable AI systems.