The Mysterious World of Deepfake Technology

Deepfake technology can flawlessly merge anyone from anywhere in the globe into a movie or photograph in which they never took part. It also includes voices, like in an altered video, Facebook's Mark Zuckerberg brags about having billions of people's data, or Belgium's Prime Minister links the Coronavirus outbreak to climate change during a manipulated recorded address.

Artificial intelligence effectively learns how to transplant a source face onto a target, generally an actor, as if it were a mask, by learning how to look at it from various angles. The use of generative adversarial networks (GANS) to throw two AI systems against each other, one creating fakes and the other analyzing them, taught the synthesis engine to make better forgeries, leading to significant development.

How Do You Spot A Deepfake?

It has become increasingly difficult, as technology develops. According to US specialists, deepfake faces do not blink regularly. Because most images show people with their eyes open, the algorithms never really learn about blinking. It appeared to be a solution for the detection problem at first. But, as soon as the study was published, deepfakes with blinking lights surfaced.

Deepfakes of poor quality is simpler to identify. It's possible that the lip-synching is off, or that the skin tone is uneven. Deepfakes have a particularly difficult time portraying small features like hair, especially where strands are visible on the fringe. Strange lighting effects, such as irregular illumination and reflections on the iris, can potentially be a clue, as can poorly portray jewellery and teeth.

Deepfake detection research is being funded by governments, institutions, and tech corporate bodies. Microsoft, Facebook, and Amazon sponsored the first Deepfake Detection Challenge. It will involve research teams from around the world fighting in the game of deepfake detection for masters.

How Dangerous Are Deepfakes?

Deepfakes now pose a major threat to women, with nonconsensual pornography accounting for 96 per cent of deepfakes spread on the Internet. Deepfakes are increasingly being used to make fake revenge pornography, according to reports. 

However, bullying is not only limited to women, anyone can be placed into ridiculous, dangerous, or compromising scenarios in schools and workplaces. Deepfakes can bring profound emotional and mental pain to the person involved, as well as potentially breaking copyright laws.
Deepfakes most commonly target politicians and celebrities.

Computer scientists, of the University of Washington, have used neural network AI to predict the form of Obama's mouth and make it lip-sync. Mostly High-profile figures are used since their public profiles provide abundant source material for AI to learn from. However, technological advancement allows using anyone face.

Deepfake technology's serious effect is likely to be political. In 2018, Gabon's President, Ali Bongo, who had been suspected of being ill for some time, appeared in a strange video to reassure the population, sparking an attempted coup. It's easy to make a deepfake film that appears to depict a world leader waging war on a foreign country. This is especially problematic now that AI-generated voices that closely resemble the target have advanced.

Does The Government Have The Right To Ban Deepfakes?

Almost certainly not, because the government is prohibited by the First Amendment from restricting speech solely because it is incorrect. Although social media sites are not bound by the First Amendment, they may want to mark deepfakes as such rather than removing them outright in situations where people may be misled.

Twitter is now putting a notice on tweets containing manipulated media or deepfakes, indicating to viewers that the media is manipulated. Users who plan on retweeting, like, or engaging with the tweet will see an additional warning.

Deepfakes In India

During the Delhi Assembly elections in February 2020, a video showing Delhi's BJP President criticising his opponent leaked on the internet. This video was created by combining an older video of his with another in which he was discussing an entirely different topic. This was one of the first times deepfake technology was used in an Indian election.

However, there is no explicit law against deepfakes in India. Sections 67 and 67A of The Information Technology Act 2000, among other laws, allow for the punishment of posting sexually explicit material in electronic form. Defamation is punishable under Section 500 of the Indian Penal Code 1860, but these rules are insufficient to deal with the many varieties of deepfakes.

In a recent response to deepfake technology, India's IT minister anticipated that deepfakes would only be used to spread fake news. Deepfakes, on the other hand, can be used merely for fun while still invading someone's privacy. 

As a result, in addition to raising public knowledge about this unique technology, the government should pay close attention to the issues posed by deepfakes before they become a threat in India.

Written by - Kriti Verma
Edited by - Piyush Pandey