DEEP FEATURE-BASED IMAGE FORGERY DETECTION USING THE RIFD-NET FRAMEWORK

Abstract

The authentication of image originality has gained prominence owing to the proliferation of social media and digital photography. Journalism, legal evidence, and cybersecurity are particularly vulnerable to image forgery, which is defined as the alteration or modification of digital photographs. Conventional forgery detection methods sometimes fail against complex editing strategies, exhibiting deficiencies in both resilience and accuracy. In addressing these issues, the authors provide RIFD-NET, a deep learning architecture proficient in properly identifying various forms of picture forgery. RIFD-NET utilizes a multi-branch convolutional neural network design to detect subtle forging artifacts, using feature extraction from both spatial and frequency domains. Attention methods are used to improve detection effectiveness under challenging settings like as compression, noise, and scaling by concentrating on modified areas. The suggested paradigm may successfully generalize several types of forgeries, including copy-move, splicing, and object removal attacks. RIFD-NET surpasses current leading forgery detection systems in accuracy, precision, recall, and robustness, as shown by experimental assessments on benchmark datasets. Empirical evidence from ablation experiments confirms the effectiveness of attention and hybrid feature extraction modules. Moreover, RIFD-NET demonstrates computational efficiency appropriate for real-time applications and indicates considerable generalizability for novel forging strategies. The suggested architecture provides a strong, scalable, and dependable method for verifying digital media authenticity and conducting picture forensic analysis.