Deepfakes: Revealing the Foundations of Deceptive Machine Learning

December 20, 2020

Deepfakes: Revealing the Foundations of Deceptive Machine Learning

Advances in deep learning have given rise to a disruptive and potentially devastating capability, deepfakes. The term “deepfake” refers to media (images, video, audio, and text) that are either partially or fully-synthesized, but are remarkably realistic and believable. This talk will provide an overview of the underlying technologies behind deepfakes, their history and progression, positive and negative applications, and finally their ethical and societal implications.

About the speaker:

Dr. Robert Wright is a Principal Research Scientist with Assured Information Security (AIS) and group lead from the Advanced Research Concepts team. He serves as the Principal Investigator (PI) for multiple DARPA and AFRL efforts including the VIC3TORS effort that is developing novel machine learning-based biometric verification techniques for cyber attribution. In 2014, he received his PhD in computer science from Binghamton University for his work in developing reinforcement learning algorithms that learn efficiently from experience. His work has been published in many top tier scientific venues and was awarded “Best Paper” at the 2013 European Conference on ML.

Prior to AIS, Dr. Wright was a Computer Scientist with the Air Force Research Laboratory (AFRL) and co-lead for the Autonomy Community of Interest: Machine Perception Reasoning and Intelligence technical challenge area for the Office of the Secretary of Defense. In his 13 years with AFRL, he led several research projects in machine learning and autonomous systems. His research interests include reinforcement learning, machine perception, generative adversarial networks, multi-agent systems, and genetic algorithms.