RIFD-NET: A ROBUST IMAGE FORGERY DET ECTION NETWORK

Abstract

With the widespread use of digital imaging and social media platforms, the authenticity of images has become increasingly critical. Image forgery, which involves manipulating or tampering with digital images, poses significant threats in various domains including journalism, legal evidence, and security. Traditional forgery detection methods often struggle with robustness and accuracy, especially against sophisticated editing techniques. To address these challenges, we propose RIFD-NET, a novel deep learning-based framework designed to detect various types of image forgeries with high precision and resilience. RIFD-NET leverages a multi-branch convolutional neural network architecture that integrates both spatial and frequency domain features. This hybrid feature extraction approach enhances the model’s ability to identify subtle inconsistencies introduced by forgery processes. Additionally, the network incorporates attention mechanisms to focus on manipulated regions, improving detection performance even under adverse conditions such as compression, noise, and scaling. The design allows RIFD-NET to generalize effectively across diverse forgery types including copy-move, splicing, and removal attacks. In conclusion, RIFD-NET represents a significant advancement in the field of image forgery detection by delivering enhanced robustness, accuracy, and adaptability. Future work will focus on extending the framework to video forgery detection and incorporating explainability features to provide better interpretability of detected manipulations. The proposed network offers a promising direction for strengthening trust and authenticity in digital visual content.