System for Deepfake Detection - DeepAudioDetector - Ekoparty 2021
Ekoparty Security Conference via YouTube
Overview
Explore a comprehensive conference talk from Ekoparty 2021 on deepfake detection systems, focusing on the DeepAudioDetector project. Delve into the challenges posed by digital information overload and the evolution of content alteration through AI-powered deepfakes. Learn about two research projects aimed at detecting deepfakes, including methodologies, experiments, and evaluation techniques for detection tools. Discover a freely available tool for deepfake detection and gain insights into future work in this field. The presentation covers topics such as unsupervised learning, reinforcement learning, adversarial machine learning, data science, audio deepfakes, data processing, and experimentation results. Benefit from the expertise of speakers Christian Camilo Urcuqui López, Kevin Zarama, and Cristhian Eduardo Castillo Meneses as they share their knowledge on cybersecurity, data science, and artificial intelligence.
Syllabus
Introducción
Aprendizaje no supervisado
Aprendizaje por refuerzo
Adversarial Machine Learning
Ciencia de datos
Resultados
Audio Deepfake
¿Es realmente un problema?
Procesamientos de los datos
Experimentación
Modelo base
Experimento 2
Conclusiones
Trabajo a futuro
Repositorio del Proyecto
Taught by
Ekoparty Security Conference