Image Forgery Detection Using CNN

Image forgery detection has become increasingly important in the digital age due to the rising misuse of editing tools that enable tampering with images. This poses significant threats to fields such as media, law, and forensic investigation where authenticity is vital. Effective forgery detection techniques are therefore necessary to ensure image integrity and trustworthiness. Forgery types such as copy-move and splicing introduce subtle artifacts that are often undetectable to the human eye. To detect such manipulations, Convolutional Neural Networks (CNNs) are employed due to their ability to learn hierarchical features. This work proposes a CNN- based framework using MobileNetV2 and transfer learning for classifying forged and authentic images. The approach utilizes Fdiff images—derived by calculating pixel-wise differences between original and compressed versions—to highlight tampering artifacts. These are processed through a pre-trained network for accurate classification. The proposed system is implemented using Python and TensorFlow and demonstrates high accuracy while maintaining low computational cost, making it suitable making it deployable in practical real-world environments.