Some time ago, a deep fake video of a movie actress called Rashmika Mandana went viral. The video had her face morphed over a British Indian influencer’s video. Within days of this incident, an indecent AI generated deep fake image of Bollywood star Katrina Kaif also went viral. Such incidents are of course unacceptable. I wanted to take this opportunity to share widely the resources available to deal with such incidents. For any cases where inappropriate pictures are uploaded, one can create a case on the website stopncii.org. This organization will hash the image, scrub the internet, and take down matches of the image found in all partner websites. They also have a helpline and case tracking.

While, cyber bullying and a sextortion etc. are not strictly inside my wheelhouse, there are obvious ways the same tools can be used in cyber fraud as well. I came across at least two videos this week. In one video, there was an app called Xpression Camera used to create a talking photo deep fake in a typical catfishing scheme. But then again romance scams don’t even need such elaborate productions. Even without a deep fake, one delusional lady thought she was in a relationship with Mark Zuckerberg. Yup, that same one.

There are several tell tales that help spot a deepfake. These are –

• Strange shadows that should not exist
• Poor lip syncing
• Contextual inconsistencies such as different earrings
• Weird teeth (sometimes one too many)
• Unnatural intonations
• Blinking pattern may seem odd
• Tools for detecting fakes

At least two tools that I am aware of, to detect deep fakes were these –

• Aiornot.com
• Wasitai.com

If we think such fraud is not happening in India then here’s a story from Ludhiana for you. A businessman in this sleepy town received a WhatsApp call from a ‘client’ from an unknown number. The caller claimed that his old number had been discontinued and urgently wanted Rs. 10 lac. The voice was mimicked, and so it didn’t raise any suspicion as such dealings were common between the two. A little research can show multiple such free tools where typed text can be spoken by the voice by providing very limited sample to these AI tools. Obviously, I’m trying not to share fraud perpetration tools, while focusing on fraud detection tools. The Indian government has also helpfully sent a timely reminder to social platforms that it’s 3 years jail time and Rs 1 lac in fines for deepfake crimes. One could argue there are completely non-criminal uses for a deepfakes. This one time, I was driving out on a holiday, but had one last call that I had to take before heading out. Decided to take it while driving. You know how some drugs come with the warning do not drive or otherwise operate heavy machinery. This meeting should have provided that advisory. A very ungifted speaker from HR droned on for like an hour. Now there’s a great use case for the Xpression Camera app 😊 Not suggesting any of you use it for skipping meetings, or anything like that.