To uncover a deepfake video call, ask the caller to turn sideways

#1

Post by Hard_Space » Mon, Aug 08, 2022, 12:26 PM UTC

Re: To uncover a deepfake video call, ask the caller to turn sideways

#2

Post by roessland » Mon, Aug 08, 2022, 1:45 PM UTC

Might be a great article but I had to stop reading since I couldn't bear the scroll hijacking.

#3

Post by budafish » Mon, Aug 08, 2022, 1:47 PM UTC

roessland wrote: post #2

Might be a great article but I had to stop reading since I couldn't bear the scroll hijacking.

100% agree. Made me feel a bit nauseous.

#4

Post by mdp2021 » Mon, Aug 08, 2022, 1:48 PM UTC

roessland wrote: post #2

Might be a great article but I had to stop reading since I couldn't bear the scroll hijacking.

No issue here. It appears your system allows it.

#5

Post by bobkazamakis » Mon, Aug 08, 2022, 1:48 PM UTC

shoe on head

#6

Post by diydsp » Mon, Aug 08, 2022, 1:49 PM UTC

8/9/2022: To prevent against uncovering, train your models to generate sideviews.

#7

Post by david_draco » Mon, Aug 08, 2022, 1:49 PM UTC

"profile view challenge" coming in 3, 2, 1 ...

#8

Post by nominusllc » Mon, Aug 08, 2022, 1:49 PM UTC

roessland wrote: post #2

Might be a great article but I had to stop reading since I couldn't bear the scroll hijacking.

I did not experience this, my system doesn't allow it

#9

Post by shiftpgdn » Mon, Aug 08, 2022, 1:52 PM UTC

Tangentially related but a simple way to bust a chatbot is to ask “What is larger, the Eiffel Tower or a shoe box?”

#10

Post by tommoor » Mon, Aug 08, 2022, 1:52 PM UTC

diydsp wrote: post #6

8/9/2022: To prevent against uncovering, train your models to generate sideviews.

Yes, but one of the points of the article is a distinct lack of source material to train models on profile views