A robust misinformation-aware image forgery detection framework using a Pseudo-Label Clifford Graph Fusion Convolutional Neural Network with Goat Optimization

The rapid spread of manipulated images across digital platforms has increased the need for reliable Image Forgery Detection (IFD). This study proposes a misinformation-aware IFD framework based on a Pseudo-Label Clifford Graph Fusion Convolutional Neural Network with Goat Optimization (PCGCNN-GO). The framework combines FGDIF for preprocessing, UADT for uncertainty-aware feature extraction, PSFAN and CSCNN for graph-based and geometric feature learning, and GO for parameter optimization. Experiments were conducted on the CASIA 2.0 and MICC-F2000 benchmark datasets. PCGCNN-GO achieved 98.92% accuracy and 98.65% precision on CASIA 2.0, compared with 94.67% accuracy for the best listed comparison method. On MICC-F2000, it achieved 99.08% accuracy and 98.80% precision, compared with 95.12% accuracy for the best listed comparison method. The framework also required lower computational time than the listed comparison approaches. These results demonstrate the effectiveness of the proposed integrated strategy for accurate and efficient image forgery detection.