An Ensemble Approach to Image Tampering Detection: Methods and Analysis

Abstract

Image forging or manipulation, spread nowadays, is a serious problem digital image processing faces. As well as maliciously altering personal pictures, it covers spreading misinformation. So, in this respect, such a model that is designed to monitor unwanted alteration of images is needed nowadays more than ever.• The proposed methodology introduces a novel method based on deep learning, made specifically to detect image manipulation. Due to this algorithm's ability to spot photo manipulated areas, any unauthorized changes can be quite precisely identified.• The methodology proposed involved both the use of Columbia dataset and Synthetic dataset.• Test results indicated that the model does very well on many critical performance parameters, for example, accuracy, precision, recall, and F1-score. Experiment further gauged the model performance using real-world photos to confirm if it was reliable and applicable in real world conditions.• In summary, the proposed method for picture manipulation can be used in the best way to detect unauthorized photo adjustments. The approach can further be applied to media forensics in detecting false information and confirming the authenticity of photographs.

Keywords: Image Tampering, Forgery Detection, Deep Learning, feature extraction, Yolov7, Convolution Neural Network