China to rein in deepfake tech starting next month

China is working to more tightly scrutinize so-called “deepfake” technology and services next month as unchecked use of deepfakes could lead to its use in criminal activities.

Details:

• China has been pushing for new synthetic content regulations.
• Earlier this year, China released “Provisions on the Administration of Deep Synthesis Internet Information Services”.
  • It was basically a draft of regulations for deep synthesis technology, an umbrella term covering “text, images, audio, video, virtual scenes, or other information” created with generative models.
• The new regulations are aimed at:
  • Deep synthesis service providers and emphasize cybersecurity
  • real-name verification of users
  • data management
  • marking of synthetic content to alert viewers
  • Dispelling rumours

What is a deepfake?

• Deepfakes use a form of artificial intelligence called deep learning to make images of fake events, hence the name deepfake.
• This new technique allows unskilled people to make deepfakes with a handful of photos, and fake videos.

Is it just about videos?

• The answer is ‘No’. Audio can be deep faked too, to create “voice skins” or” voice clones” of public figures.

How are they made?

• First running thousands of face shots of the two people through an AI algorithm called an encoder.
• The encoder finds and learns similarities between the two faces, and reduces them to their shared common features, compressing the images in the process.
• A second AI algorithm called a decoder is then taught to recover the faces from the compressed images.
• Another way to make deepfakes uses what’s called a generative adversarial network, or GAN.

Why are deep fakes becoming harder to detect?

• This type (GAN) of machine learning system consists of two neural networks, operating in concert. One network generates the fake and the other tries to detect it, with the content iterating back and forth, and improving with each iteration (repetition of the process).
• This dynamic is replicated in the wider research landscape, where each new deep fake detection technique gives the deep fake makers a new challenge to overcome, making deep fakes increasingly foolproof.
• Thus, any deep fakes detector will only be going to work for a short while before deep fake makers account for it in their algorithm.

What technology does one need?

• High-end desktops with powerful graphics cards
• Several companies are doing all the processing in the cloud.

How are deepfakes detected currently?

• Currently, deep fakes are identified manually or by software, using some identifiers like:
  • Flicking, blur with bleeding color, etc. in poorly produced deep fake videos
  • Unusual eye blinking pattern in deep fake videos
  • Using markers known as “soft biometrics” of a person i.e., his/her eyebrow movements, lip movements, etc.

Dangers Associated with Deep Fake

• New Front of Warfare: A deepfake could act as a powerful tool by a nation-state to undermine public safety and create uncertainty and chaos in the target country.
• Targeting Women: The malicious use of a deepfake can be seen in pornography, inflicting emotional, reputational, and in some cases, violence toward the individual.
• Damage to Personal Reputation: Deepfake can depict a person indulging in antisocial behaviour and saying vile things.
• Undermining Democracy: A deepfake can also aid in altering the democratic discourse and undermine trust in institutions and impair diplomacy.
• Disrupting Electioneering: A deepfake of a political candidate can sabotage the image and reputation of a contestant.

Steps taken:

• Governments, universities, and tech firms are all funding research to detect deepfakes.
• Deepfake Detection Challenge: It is backed by Microsoft, Facebook, and Amazon.
  • It will include research teams around the globe competing for supremacy in the deepfake detection game.
• Facebook has been banning deepfake videos that are likely to mislead viewers into thinking about someone.
• Certain US states like New Jersey and Illinois have introduced local privacy legislation that addresses deepfakes.
• Major US platforms like Facebook and Twitter have created new systems meant to detect and flag deepfakes.