Annual General Meeting (AGM) - Deepfake Videos - Generation and Detection

When:  May 16, 2022 from 11:30 to 13:00 (CT)
Associated with  Huntsville Chapter
Topic:  Deepfake Videos - Generation and Detection
Speaker: Dr. Catherine Bernaciak & Dr. Shannon Gallagher, Software Engineering Institute - Carnegie Mellon (SEI - CMU)
CPE Hours: 1

Background:

A deepfake is a piece of media—either image, video, or speech, typically representing a human subject—which has been altered to be deceptive using deep neural networks (DNNs) and involves the substituting of identities between a source and a target subject. In this talk, we describe the process of Deepfake creation and discuss state-of-the art detection methods for identifying deepfakes in the wild.


Lessons Learned:
We learned two things about the creation of deepfakes: (1) They can be done with relatively inexpensive GPU hardware (gaming-type GPUS) (2) Significant post-processing skills and voluminous amounts of training data (videos) are required to make a 'high-quality' deepfake. therefore it is not as simple to make a high-quality Deepfake as one may be led to believe by media reports, however, with the right data snd skillset and a few thousand dollar hardware investment, it can be done. The creation models are advancing rapidly and with the right skillset extremely believable deepfakes can be made. Detection is a cat-and-mouse game, as the same technology used to create then can be used for detect them, and vice-versa.

Learning Objectives:

  • High-quality deepfakes are not difficult to make but require extensive skill and subject data.
  • Detection techniques are similar to creation techniques in terms of neural network architecture.
  • Detection and creation are a cat-and-mouse game.

Speaker Bio:
 

Dr. Catherine Bernaciak is a research scientist in the CERT Division at the Software Engineering Institute at CMU. She performs research in machine learning and quantum computing and has a background in Physics and Data Science. She graduate from SUNY University of Buffalo with a Ph.D. in Physics and has been with SEI-CMU since 2019.

Speaker Bio:
Dr. Shannon Gallagher is a data scientist in CERT at the Software Engineering Institute. She collaborates with other researchers to work on machine learning and statistical problems including code weakness identification and classification of images. Dr. Gallagher received her PhD in Statistics in 2019 from the Department of Statistics & Data Science at Carnegie Mellon University. Additionally, Dr. Gallagher has experience with biostatistical analyses from her work in graduate school in addition to her post doc at NIH. Besides enjoying developing and implementing statistical models, Dr. Gallagher also has a passion for data visualization. She always welcomes discussion with potential collaborators.

_past_events

Location

The Atrium at Redstone Federal Credit Union
220-A Wynn Drive
Huntsville, AL 35893
Online Instructions:
Url: http://us06web.zoom.us/j/89655552550?pwd=K0NXOXp4NXFraHpiVzY5MWVteHpFZz09
Login: Meeting ID: 896 5555 2550 Passcode: 462825

Pricing Information

Registration Price
All Registrants $0.00
   
RSVP In-Person Meal (Meat) Price
All Registrants $0.00
   
RSVP In-Person Meal (Veggies) Price
All Registrants $0.00