Human vs. AI: Leveraging Machine Learning and Deep Learning to Verify Image Authenticity

The sophistication of generative AI models has led to AI-generated images that are nearly indistinguishable from human-made ones, presenting challenges in digital art, media, and content verification. This study addresses these challenges using both traditional machine learning (ML) and deep learning (DL) approaches. We leverage a balanced dataset from Hugging Face, containing equal numbers of human-created and AI-generated images, and use Gray Level Co-occurrence Matrix (GLCM) features and Local Binary Patterns (LBP) for texture analysis to discern subtle differences. Evaluating models such as K-Nearest Neighbors (KNN), Random Forest, Decision Tree, Support Vector Machine (SVM), Multi-Layer Perceptron (MLP), and deep learning architectures like ResNet50 and VGG16, we find that traditional ML models with well-selected features can outperform advanced DL models in certain contexts, highlighting the importance of feature engineering. These findings offer robust tools for verifying digital image authenticity and protecting intellectual property, benefiting digital artists, content creators, forensic analysts, and online platforms.