Deep Fake Technology
Understanding Deep Fakes
In the modern digital age, seeing is no longer always believing. Deepfakes are a type of synthetic media—which includes videos, images, or audio recordings—that have been digitally altered using Artificial Intelligence (AI).
The primary goal of a deepfake is to make it look and sound incredibly realistic, making it appear as if a person said or did something they never actually did. By doing so, deepfakes dangerously blur the line between reality and digital manipulation.
The Technology Behind Deep Fakes
To understand how deepfakes are created, we must look at the specific AI tools used to build them. Deepfakes are powered by Deep Learning, which is an advanced subset of Machine Learning.
The core technology responsible for creating modern deepfakes is called Generative Adversarial Networks (GANs).
How GANs Work
A GAN consists of two separate artificial neural networks working against each other in a continuous loop:
- The Generator: Think of this as the “forger.” Its job is to study thousands of real images or videos of a person and then create a fake, synthetic video based on that data.
- The Discriminator: Think of this as the “detective.” Its job is to look at the fake video created by the Generator and compare it to real videos to spot any flaws or artificial mistakes.
Every time the Discriminator catches a mistake, the Generator learns from it and makes a better fake. This process repeats millions of times until the Generator creates a fake video so perfect that the Discriminator can no longer tell it is artificial.
Audio Manipulation
For voice cloning, deepfakes use Natural Language Processing (NLP). Advanced software analyses a person’s speaking style, tone, and pitch. Lip-syncing techniques are then used to perfectly align this fake, computer-generated audio with the mouth movements in the video.
Types of Deep Fakes and Detection
Deepfakes are commonly categorized into a few main types based on how they manipulate the original media:
- Face Swaps: Replacing one person’s face in an existing video with the face of another person.
- Voice Clones: Artificially imitating someone’s exact voice to make them say anything the creator types into a computer.
- Source Video Manipulation: Altering the original footage to make a person appear to perform actions or speak words they never did.
How to Detect Deep Fakes
While the technology is advancing rapidly, there are still ways to detect manipulated media:
- Visual and Audio Clues: Careful observers can often spot unnatural blinking, strange facial distortions, mismatched audio (where the voice does not perfectly match the lip movement), and weird lighting glitches around the edges of the face.
- Detection Tools: Major technology companies like Adobe, Microsoft, and specialized firms like Sensity AI are developing advanced deepfake detection software.
- Platform Regulation: Social media platforms are increasingly using AI to scan, flag, or completely remove malicious deepfake content before it goes viral.
Deep Fakes and the Indian Legal Framework
Currently, India does not have a dedicated, specific law meant only for deepfakes. However, law enforcement agencies use several existing legal provisions to offer protection and prosecute offenders.
Information Technology Act, 2000 (IT Act)
The IT Act is the primary law dealing with cybercrime in India.
- Section 66D: This section is used to target and punish individuals for cheating by impersonation using computer resources.
- Sections 67, 67A, and 67B: These sections are heavily used to prosecute individuals who publish or transmit deepfakes that are obscene or contain sexually explicit acts.
Information Technology (IT) Rules, 2021
These rules place strict responsibilities on social media platforms (like Twitter, Facebook, or YouTube). Under these rules, platforms are required to quickly remove impersonation or morphed content when alerted by a user. If the platform fails to remove the deepfake, it will lose its “safe harbour” protection. (Note: ‘Safe harbour’ is a legal provision that normally protects social media companies from being sued or held legally responsible for the content posted by their users.)
The Copyright Act, 1957
This act can be applied if a deepfake creator uses copyrighted images or videos of a person without their explicit permission. It legally bars the unauthorized use of content over which someone holds exclusive rights.
Institutional Framework
- CERT-In (Indian Computer Emergency Response Team): The national nodal agency for responding to computer security incidents. It regularly issues public advisories on deepfake threats and safety measures.
- I4C (Indian Cybercrime Coordination Centre): A government body that actively assists local law enforcement agencies across the country in effectively investigating and tackling complex cybercrimes, including deepfake networks.
Landmark Judicial Interventions in India
Indian courts have stepped in to protect individuals from having their digital identities stolen or manipulated by AI. These cases focus heavily on the concept of Personality Rights (the right of a person to control the commercial use of their name, image, likeness, or other identifying traits).
Anil Kapoor’s Case (2023):
- The Delhi High Court granted an ex-parte (in the interest of one side) injunction completely restraining unauthorized individuals from using Bollywood actor Anil Kapoor’s name, image, or famous dialogue phrases for commercial gain through AI or morphing. The court noted that a person’s personality rights deserve strict protection to maintain the dignity of the individual and their family.
Mr. Shivaji Rao Gaikwad (Rajnikanth) vs M/S. Varsha Productions (2015)
- The Madras High Court granted an injunction restraining the use of actor Rajnikanth’s name, image, caricature, and unique dialogue delivery style in a film. This was a landmark moment in formally recognizing and protecting the personality rights of a celebrity against unauthorized imitation.
