Combating Digital Forgeries: Advanced AI Techniques for Detecting Forgeries of Diverse Data

Detecting forgeries in diverse datasets involves identifying alterations or manipulations of various digital media, such as images, videos, and documents. There are unique challenges associated with each type of media for efficient forgery detection. This study investigated the effectiveness of both traditional and AI-based feature extraction techniques, combined with machine learning classifiers, in detecting forgeries across three distinct datasets: document images, the CASIA 2.0 standard image forgery dataset, and a video dataset containing real and forged videos. Traditional feature extraction methods such as Local Binary Patterns (LBP), Gray-Level Co-occurrence Matrix (GLCM), color histograms, and RGB to HSV conversion were utilized alongside machine learning classifiers like Support Vector Machines (SVM), Random Forest (RF), and Gradient Boosting Machine (GBM). Additionally, AI-based techniques leveraging pretrained models like ResNet50, VGG16, EfficientNetB0 and ensemble methods with Principal Component Analysis (PCA) were employed. This study evaluated the performance of these methods using metrics such as precision, recall, F1 score, and accuracy across different datasets. The results demonstrate the effectiveness of using ensembled features with PCA in improving the detection capabilities of the models. This study underscores the importance of leveraging traditional and advanced AI methodologies to effectively combat digital forgeries in various media types.