Deep Fake Detection

The process of producing deepfakes has become much easier and quicker, which allows one to make rather convincing fake photos and videos. These may pose a major danger to integrity, privacy and security of our digital lives. Advanced changes are difficult to identify with traditional forensic tools such as metadata analysis and manual examination when artifacts have either been subtly or inconsistently dated. The system employs well chosen sets of real and fake images and videos of the various types of face changes, including both, features, which are spatial and temporal, and those obtained in a variety of frame sequences. The images that are uploaded are resized, standardized and undergo an organized workflow. They are divided into videos, face detection, and standardized 112112 transformations and then sequence is formed. They are two convolution-based classifiers, a simple CNN2D and an adapted model, which has more convolutional block and more effective regularization. The performance is judged based on standard measures and the modified CNN2D scored the highest 95.775 percent, which was higher than the baseline method. The approach enhances its detection capability of manipulation by using optimized preprocessing and powerful architectures alongside multi-mode analysis to ensure that the fake visual information is rightly detected. “