Deep Learning Framework for Robust Deep Fake Image Detection: A Review
Ankita Bhandarkar, Prashant Khobragade, Raju Pawar, Prasad P. Lokulwar, Pranay Deepak Saraf · 2024
Deepfake technology, which uses deep learning to change pictures and videos, makes it difficult to identify and tell the real and fake images. People are worried about false information, data breaches, and bad uses of deepfake generation methods that are getting better and better very quickly. As the name suggests, this study looks at the current deep learning methods for finding deepfake pictures, focusing on their pros and cons. It looks at how well CNN method, autoencoders method, RNN method and attention-based models can discover little visual blemishes and blunders that were included amid the deepfake creation handle. The study also talks about how data addition, transfer learning, and ensemble learning can help improve the efficiency of identification. Additionally, it talks about how deepfake technology is changing and the need for flexible models that can successfully fight more complex fakes. Problems like small datasets, heavy processing needs, and the chance of hostile attacks are talked about, along with possible answers such as combining blockchain technology and real-time detection systems. This review gives an indepth look at the latest progress and gaps in deepfake detection. The goal is to help make tools that are stronger, more scalable, and more reliable to fight the growing threat of deepfake picture manipulation.