부산대학교 도서관

In the digital era, the proliferation of image manipulation tools has led to an alarming surge in the creation of spurious images capable of misguiding and deceiving viewers. These fabrications encompass a diverse spectrum of manipulations, encompassing techniques such as image splicing, copy-move operations, and facial modifications. To address this growing challenge, this research paper delves into the domain of deep learning, a cutting-edge technology renowned for its ability to decipher complex patterns in data. The primary objective is to improve underlying mechanisms of the existing techniques such as CNN, GAN, Transfer Learning and Edge Feature Utilization. By providing insights into the capabilities and limitations of deep learning techniques, this study lays the groundwork for the development of more precise and efficient approaches to address the challenges posed by counterfeit images. The experimental finding demonstrates that proposed modifications exhibited an average accuracy improvement of $6 \%$ and a $5 \%$ increase in F1-score when contrasted with existing methods.