#180: Deepfakes | One stop resource for UPSC
6 minutes read, all points covered
Rashmika Mandhana case brings a critical problem to surface: DEEPFAKE
While law takes its course, it is important to learn all dimensions of this case
Deep Synthesis
• Use of tech to generate text, images, audio/video, to create visual scenes
• It uses deep learning and augmented reality
• Deepfake is an application of this tech
What is a Deepfake?
Deepfakes use synthetic media to swap face/voice of one person
It compiles artificial media with Machine Learning algorithm
It has the power to ‘manufacture’ people who do not exist OR..
make existing people say things they did not say!
How does it work?
Hidden layers in neural network replicate human brain learning
Then these layers perform mathematical transformations
These convert input signals to output signals
These signals become fake media
Where all are Deepfakes being used?
Rashmika’s case is not the first one
In the past, we have seen deepfakes of leaders like:
Obama, Trump, Modi, Zuckerberg (and many more!)
Where did this term come from?
2017: Anonymous Reddit user manipulated Google’s algorithm
What for?
To create Pronographic videos via face swapping
Catch here is, celebrity faces were swapped to create their porn!
What are other applications?
• Scam & hoaxes
• Social Engineering
• Election Manipulation
• Celebrity Pornography
• Identity theft & Financial Frauds
• Automated disinformation attacks
Looking at present case,
several voices including Bachchan called for legal route
At present, China has strongest policy in place
What is it?
• Any doctored content to be explicitly labeled
• Any such content shall be tracked back to its source
• Seek consent of any person whose image/voice is being edited
• Any such news can only be from govt-approved list of news outlets
• Such service providers MUST abide by local laws, respect ethics
• Further, they shall maintain correct political direction & public opinion orientation
How are other countries dealing with it?
a) European Union
• Published Code of Practice to stop spread of disinformation
• Tech giants shall take measures to counter deep fakes
• Non compliance leads to fine to tune of 6% of global turnover
b) United States
• Bipartisan Deepfake Task Force Act
• Annual study to be conducted to tackle any arising issues
• Respective states have specific laws
What are safeguards in India?
• Section 67 of IT Act
Used to check obscene material in e-form online
Other than these, there are some IPC safeguards
However, these are not enough!
How to safeguard yourself?
• Avoid downloading lesser known apps
• Disallow all immaterial app permissions
• Audit all app permissions on your phone
• Disable use of voice assistants like Alexa/Siri
• Disable face scan lock on phone, use passwords
• Create media backup on Pen drive/HDD (not cloud)
• Don't be shy of registering legal complain if you come across any such media against you
Idea is to minimize training algorithm on your data
It is hard but worth it!
Way Forward (for mains answers)
Rashmika Mandhanna episode highlights the issue
It is not happening for the first time and India needs to act fast
• Set up an expert panel or task force to study the issue in time bound manner
• Ensure stricter rules in under construction Bhartiya Nyay Samhita
• Sensitize the citizens to be alert about any such media
• Involve Big Tech companies on stricter terms
• Set up stronger processes around VPN use
• Encourage spirit of Article 51A(h)
What else?
