---
title: "Discussion on \"From Zero to Deepfake Detector: How I'm Building Vigil AI to Protect India From the Coming AI Fraud Wave\" | Hashnode"
description: "Discussion on \"From Zero to Deepfake Detector: How I'm Building Vigil AI to Protect India From the Coming AI Fraud Wave\". subtitle: \"One developer. Zero funding. The most important cybersecurity problem in India right now.\" slug: vigil-ai-deepfake-detector-india-startup-journey tags: startup, ai, machinelearning, python,"
url: "https://hashnode.com/posts/from-zero-to-deepfake-detector-how-i-m-building-vigil-ai-to-protect-india-from-the-coming-ai-fraud-wave/69f2c7fc6e0124c05e96f652"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [From Zero to Deepfake Detector: How I'm Building Vigil AI to Protect India From the Coming AI Fraud Wave](https://vigilai.hashnode.dev/from-zero-to-deepfake-detector-how-i-m-building-vigil-ai-to-protect-india-from-the-coming-ai-fraud-wave?utm_source=hashnode&utm_medium=feed)

subtitle: "One developer. Zero funding. The most important cybersecurity problem in India right now." slug: vigil-ai-deepfake-detector-india-startup-journey tags: startup, ai, machinelearning, python,

Apr 30 · 11 min read · vigilai.hashnode.dev  
Read the article

## Written by

[Abhishek Kumar](https://hashnode.com/@abhishek-nine)

@abhishek-nine · bhilai , india

  
[View profile](https://hashnode.com/@abhishek-nine)

AI developer building www.vigilai.online

## Responses 2

Markdown supported · ⌘↵ to post

[Abhishek Kumar](https://hashnode.com/@abhishek-nine)

[Apr 30](https://hashnode.com/posts/from-zero-to-deepfake-detector-how-i-m-building-vigil-ai-to-protect-india-from-the-coming-ai-fraud-wave/69f2c7fc6e0124c05e96f652/comment/69f2c8386e0124c05e971def)

love it

[Jorge Demetrio Sobieski Teixeira](https://hashnode.com/@jorgedemetrio)

[Jun 29](https://hashnode.com/posts/from-zero-to-deepfake-detector-how-i-m-building-vigil-ai-to-protect-india-from-the-coming-ai-fraud-wave/69f2c7fc6e0124c05e96f652/comment/6a42b20c922b2f10d9478138)

Great initiative, Abhishek. One angle worth baking in early: the most robust signal usually is not in perception (blinking, lip-sync) but in the frequency domain. GANs and diffusion leave a periodic upsampling fingerprint in the spectrum that a 2D FFT plus a radial average collapses into a single score. Pairing that with Error Level Analysis for splicing and a face-landmark check (MediaPipe) gives you an ensemble that degrades gracefully instead of one brittle model. I put together a Python walkthrough of exactly this ensemble over on Meu Universo Nerd if it helps your build.
