Overview
Explore the world of deep fake detection in this Black Hat conference talk. Delve into the dangers of neural network-generated speech and its potential for social engineering attacks. Learn about the history of text-to-speech technology, GAN discriminators, and the innovative approach of using mice for deep fake detection. Discover the biological aspects of speech perception, including coarticulation and synaptic processes in mice. Gain insights into the challenges of creating convincing deep fakes and the importance of developing effective detection methods. Understand the intersection of machine learning and biology in the fight against audio impersonation attacks.
Syllabus
Introduction
Todays Fakes
The Dangers
Informal Study
Machine and Biology
What is a Deep Fake
Making Good Deep Fakes
Wavenet
History of TTS
Gann Discriminator
References
What Kind of Deep Fake Detection
Listening to Speech
CoArticulation
Deep Fakes
Mice
Mouse
Why Mice
synapses
reset synapses
local inhibitory neurons
summary
Taught by
Black Hat