Comprehensive Analysis of Deep Fake Detection Techniques for Images and Videos using Deep Learning

Deep fake detection is the process of identifying and recognizing manipulated media such as videos or images, which replicate original people through Artificial Intelligence (AI) techniques. AI methods like Machine Learning (ML) and Deep Learning (DL) approaches aim to differentiate among reliable and altered contented to avoid misinformation. Distinguishing between original and forged images is difficult due to the similarity in their features, often leading to misidentification and causing inaccurate detection. This survey aims to analyze deep fake video and image detection methods and highlights the many challenges encountered during the detection of deep fake content. Detection approaches for videos include Dense Swim Transformer Net (DST-Net) and Directional Magnitude Local Hexadecimal Pattern (DMHLP), while image-based detection approaches include Vector Auto Regressive Moving Average (VARMA), and Gated Recurrent Unit (GRU), referred to as VARMA-LSTM-GRU. Performance metrics such as precision, F1-score, and accuracy are utilized to evaluate deep fake detection models.