Deep Fakes are regularly in news with the likes of former POTUS Donald Trump, PM Modi etc. being subjected to test.
I’d keep updating this post on deep fakes. To begin with, here is all you need to know about Voice Deep fakes
What are voice deep fakes?
One that closely recreates mimics voice of a real person
It can replicate the characteristics, tonality, phonetics etc. of one’s voice
How are they created?
Generated using AI & robust computing power
Powerful graphic cards in high-end computers needed to leverage cloud computing
The training data is fed to AI models
This data is often the recorded samples of target’s speech(es)
It can take weeks to produce such ‘fakes’
Ethical Issues and Threats wrt Voice Deep Fakes
Identity theft and frauds
Fake news and phone scams
Unethical use in filmmaking, porn and related industries
Some examples
A bank manager in UAE was called via a deep fake of a company’s director authorizing $35mn transfer
A UK-based CEO was directed by a deep fake of a Hungarian based supplier to transfer $243k
Tools to clone a voice
OpenAI’s Vall-e
ReSpeecher
Descript
Resemble
How to detect a voice deep fake?
It requires high end tech (software and hardware both) to break down speech patterns and unluckily, there are no foolproof ways yet to detect audio deep fakes. Some methods like watermarks, spam filters, multi factor authentication, anti fraud solutions etc. are in use
Hi, at @CSEWhy I invite you to support me! You can do that in three simple ways:
Become a blog contributor and have your write ups featured on my blog. Write me an email if you’d want to be a contributor at whycse@gmail.com
Share the link to this post on your WhatsApp story
-OR- Contribute a small amount (Pay What You Want) by clicking here (I send a valuable gift to everyone who makes a contribution)
To find links to my Twitter, Telegram, YouTube and Podcast Visit my Website