How to Detect Fake Faces - Manipulated Images Using CNNs
Explore cutting-edge techniques for detecting manipulated facial images using Convolutional Neural Networks (CNNs) in this 54-minute conference talk from the Hack In The Box Security Conference. Delve into the urgent need for effective methods to expose fake face images amidst rapid progress in facial manipulation technologies. Learn about two powerful approaches: a simple yet effective CNN architecture achieving 99% accuracy for Deepfakes detection, and a FaceNet-based method utilizing face recognition features to train an SVM classifier. Gain insights from AI security experts at Baidu X-Lab as they discuss how low-level and high-level features are leveraged to automatically detect facial manipulation frames. Understand the potential societal, political, and commercial implications of advanced facial manipulation techniques, and discover how CNN-based networks can effectively distinguish fake face images from real ones.