Advancing Image Forgery Detection: A CNN-Based Approach

In the digital age, image forgery is a significant concern due to advanced editing tools. This study addresses the need for reliable forgery detection, focusing on copy-move, splicing, and retouching techniques. Using Convolutional Neural Networks (CNNs) known for learning hierarchical image features, the research developed a robust detection system. The CNN model, trained on a diverse dataset which has over 10,000+ images and it is split into test and train data for better classification results, detected subtle inconsistencies indicative of manipulation, achieving a peak accuracy of 94% over 50 training epochs. Test cases showed the system effectively detected forged images and identified forgery types, though it had shortcomings with invalid inputs. The CNN architecture employed convolutional and max-pooling layers with specific hyperparameters to enhance feature extraction. This paper offers a valuable tool for digital forensics experts, law enforcement, and journalists to ensure the authenticity of visual content.