What is deepfake, and how can it steal your identity?
What if a person you watch over the internet is not a real person?
What’s new in it?
Of course, nothing new in it.
But what if there is a video where a popular person could be seen explaining something you and that person have never shot the video?
How’s that possible?
This is Deepfake for you.
What is Deepfake?
Deepfake is a synthetic media which will digitally manipulate the video or image of a person with that of another and will make it tough for the general public to understand if that is real or fake. It is the manipulation of facial appearance through deep generative methods.
Creating fabricated or fake content is not new, deepfakes have taken it to the next level by leveraging the powerful techniques from machine learning and artificial intelligence to alter or generate visual or audio to deceive easily.
It has garnered attention from across the world over the last few months as this technique could be used for creating child sexual abuse material, celebrity pornography videos, fake news, bullying, making financial fraud, fake news, revenge porn etc. This is extremely harmful or dangerous for our society.
How deepfake videos are made?
Let me make it really simple for you to understand.
The Deepfake videos are generated using two networks, an encoder and a decoder, both are part of the Generative Adversarial Network (GAN) framework. The task of the encoder network is to analyse the source content, the original face in our case, and extract crucial features and representation from that. Thereafter, it will give those features to the decoder network, which helps in generating new content, which is the fabricated content.
This continues until the AI gets its desired result. In the case of creating a deepfake video, this process takes place from each frame of the video to ensure consistency. GANs use a generator network to create a fabricated video, and then the discriminator network attempts to differentiate fake and real content.
How to identify deepfake videos or images?
Is it impossible to identify a deepfake video? Absolutely Not! Even you can identify a deepfake video or image using minor attention. Just pay close attention to video and audio inconsistencies. Here are the other ways to identify deepfake video:
- Facial Expressions
Just look closely at visual signs like inconsistent facial expressions, unusual movements, or strange-looking artefacts. You will also find out that the video frames look a little blurry or hazy. - Pay attention to psychological signs
Deepfakes can be detected by looking at physiological signals such as abnormal blinking frequency, eye colour, or heartbeat. These signs can indicate if a video is real or fake. - Biological and physiological signs
Another way to identify deepfake videos is by looking at abnormal blinking, eye colour, or lip sync. - Audio discrepancies
The deepfake audio will have unnatural pitch patterns and shifts in tone or pitch. You need to check these things whenever you doubt on video’s authenticity. - Identity information
Every person acts differently, one can find unique patterns in the speech, expression or gesture of the person in the video. These signs can help you detect inconsistencies or strange patterns, and all these can help you identify deepfake videos. - Verify the source
In the era of the internet, it has become significant to develop credibility. This is equally true in the case of deepfake videos, analyse the credibility of the video by checking its source. Ensure that the video is coming from a reliable source, not from a random YouTube.
Rashmika Mandhana’s deepfake video
The matter of the deepfake video came into the limelight when someone posted Rashmika Mandhana’s morphed video. The false video depicts Rashmika Mandhana entering an elevator. The original video featured British-Indian influencer Zara Patel, and her face was replaced with Rashmika’s face using deepfake technology.
Such incidents have become common nowadays, and have started the debate about using artificial intelligence. Technology was never a problem, but its use.
