Integrating GLCM Texture Analysis for Improved Deepfake Detection on CelebDF(v2) Dataset
Kedar Joshi, Aditya Sinha · 2024
The rapid growth in deep learning and artificial intelligence has led to the proliferation of deepfake technology, posing significant challenges in digital media authentication. This paper presents a comprehensive study on deepfake detection, utilizing Grey Level Co-occurrence Matrix (GLCM) texture analysis to enhance detection accuracy. GLCM provides critical texture features, specifically ‘Contrast’ and ‘Dissimilarity’, which offer valuable insights into the texture inconsistencies often present in deepfake images. By quantifying these inconsistencies, we aim to distinguish between authentic and manipulated images. The results from our experiments indicate that GLCM texture analysis is a powerful method for identifying deepfakes. The proposed method is evaluated on benchmark dataset, showcasing significant improvements in detection accuracy compared to existing techniques. This study adds valuable insights in the development of robust and reliable deepfake detection frameworks, offering valuable tools for protecting the integrity of digital media.