Deepfakes are a threat to India’s democracy but also video KYC: report

India is facing a proliferation of deepfakes that could spread misinformation during elections. More importantly for the biometrics industry, deepfakes could pose a major challenge for user verification, including video KYC, a new report from tech-focused media outlet Medianama warns.

The report drew its conclusions from a discussion on “Deep Fakes and Democracy” conducted by MediaNama on January 17 with support from Meta and Google. The debate centered on strategies to curb the dissemination of deepfakes and included speakers from India’s technology sector.

When it comes to identity verification, deepfakes pose a threat to video know your customer (KYC) processes which are critical for the Unified Payments Interface (UPI), the instant payment system that can be used with Aadhaar. One of the solutions proposed is biometrics.

“Biometrics are a good way to deal with a lot of these things, but once we put them in place, it doesn’t mean that they’re gonna work forever,” says Gautham Koorma, researcher, at the UC Berkeley School of Information. “There are gonna be people who are gonna break it and these things continuously need to evolve. And there is research that is showing promising results for biometrics.”

Liveness or presentation attack detection, which is used to detect deepfakes, is not explicitly mentioned in the report.

Aside from Koorma, participating in the discussion were Former Senior Director and Group Coordinator at the Ministry of Electronics and Information Technology Rakesh Maheshwari, CEO and Co-founder of policy advisory Deepstrat Saikat Datta, Co-founder of Tattle Civic Technologies Tarunima Prabhakar and more.

Deepfakes could pose an even greater threat to India’s general elections, expected to be held between April and May 2024.

The report notes that one of the reasons deepfake detection techniques have limited success is that the deepfake capabilities available to adversaries improve with each new generation. Bad actors are capable of removing watermarks designed to warn of deepfake content while audio deepfake detection algorithms become less effective when applied to content circulating on social media, despite having 90 percent accuracy in lab settings, according to the report.

There is no simple solution to deepfakes. However, companies and government agencies can implement several recommendations to reduce their impact, the report states.

Companies should make AI-generated content using their services detectable while introducing policies that regulate the creation of certain kinds of content. AI companies should alter their terms of use to ensure that they don’t monetize disinformation and misinformation spread using deep fakes. Social media platforms should also do their part by taking down or preventing their spread through shadow-banning.

The Election Commission of India could introduce a complete ban on the use of deepfakes by political parties during the upcoming general elections while digital fingerprinting can be made mandatory. Funding for deep fake detection should be increased while bolstering media literacy among the population, the report concludes.

Article Topics

biometric liveness detection  |  biometrics  |  deepfake detection  |  deepfakes  |  identity verification  |  KYC