How Network traffic patterns can be used to stop harmful videos from spreading across platforms

What is it about?

In today's world, fake and misleading videos spread quickly across various platforms like YouTube, Facebook, Instagram, X and others. Even though platforms try to stop these videos, they can be uploaded to many different platforms and slip through the cracks. Our work tries to find a way to detect and stop these videos from spreading to platforms they have not been seen on yet. We propose a system called "TripletViNet" which looks at the traffic patterns of how videos are streamed on the internet, and even though the content is encrypted, "TripletViNet" can identify a video on a new platform if it has seen it previously on another platform.

Featured Image

Photo by Alexander Shatov on Unsplash

Photo by Alexander Shatov on Unsplash

Why is it important?

Previous efforts have been able to build an AI model which trains on videos on one platform and can recognise videos on the same platform. However, it must have already seen the video on the platform it wants to recognise it on, especially because each platform streams the video differently. We remove this limitation, so we can recognise a video on a new platform (e.g. YouTube) even though we have only seen it on another platform (e.g. Facebook) by learning the differences between platforms using a special method called 'Triplet Learning'.

Perspectives