Detecting Fake Faces with AI: A Deep Neural Network Improvement Project

Abstract

Forgeries created with deep face techniques have become increasingly common in several fields in recent years, including politics, education, and the democratic process, and as a result, many scholars are working on strategies to detect and prevent such forgeries. Since these programmes typically employ machine learning or fuzzy logic, accurate data classification is not something they can promise. However, it is common knowledge that forgery detection methods necessitate a shared, massive dataset, and that facial recognition systems benefit most from deep learning’s precision. Our suggested approaches make use of images and videos from the VGG-19 shared dataset, with genetic algorithm-based feature extraction and an improved convolutional neural network handling classifications for the trained datasets, respectively. A gaussian filter is used as preliminary processing on the VGG-19 common dataset.