An Explainable Hybrid Artificial Intelligence Approach for Robust Image Forgery Detection and Authentication Using Cognitive Image Authenticity Intelligence Engine
An image manipulation develop rapidly and AI-based tools become popular among image editors allowing manipulations of high complexity and posing risks to digital forensics, journalism investigation, medicine and cyber security, the problem of image forgery detection becomes increasingly important. At present, the algorithms used in the field of image forgery detection and authentication sufferfrom the following shortcomings.It was designed to detect particular type of image manipulations, have poor generalizability to heterogeneous data sets, require significant computing resources and do not provide for explain ability of decisions made. To achieve, this research proposes to apply the concept of Cognitive Image Authenticity Intelligence Engine (CIAIE) an explainable hybrid artificial intelligence algorithm designed for the image forgery detection and authentication. The proposed algorithm utilizes such technologies as Principal Component Analysis (PCA), Conventional Neural Network (CNN), Vision Transformer (ViT), U-Net, Extreme Gradient Boosting (XGBoost), and Explainable AI (XAI) for dimensionality reduction, feature extraction, forgery localization, authenticity classification and decision making. The experimental evaluation carried out in Python showed the superior performance of the proposed approach with the accuracy of 94.68%. The proposed method was capable of achieving accurate, interpretable, and computationally efficient real-time image authentication, making it appropriate for many digital forensic and cyber security purposes.