New Tool From Jigsaw can Help Spot Doctored Images

Jigsaw, a company owned by Alphabet, has released an experimental tool to identify doctored images.

Thanks to the rise of AI-powered manipulations, fake images have become increasingly challenging to verify. Yet, the pressure to get it right is more urgent than ever, as disinformation spreads at a rapid scale and speed.

Jigsaw’s chief executive, Jared Cohen, wrote in a blog post:

“We observed evolution in how disinformation was being used to manipulate elections, wage war, and disrupt civil society. But as the tactics of disinformation were evolving, so too were the technologies used to detect and ultimately stop disinformation.”

So, the company decided to do something. On Tuesday, Jigsaw unveiled a free tool that researchers say could help journalists spot doctored images — yes, AI-generated images.

They named the tool Assembler, and here’s how it works.

Using Assembler to Spot Doctored Images

Assembler combines various existing techniques for detecting conventional manipulation methods.

The tool has seven detectors, each designed to identify a specific type of photomanipulation technique — from increased brightness to merged images. One of the detectors can also spot deepfakes created using StyleGan, an algorithm that can generate realistic imaginary faces.

All seven detectors feed into a master model that tells a user if an image has undergone alteration.

As impressive as Assembler may be, it’s not a universal remedy for manipulated media.

For one, it doesn’t address many other manipulation techniques, including those used for videos. Also, the tool exists as a separate platform, away from the social media platforms where tricksters distribute altered images.

Speaking to NYTimes, a professor at the Naples University and visiting scholar at Google AI, Luisa Verdoliva said:

“These detectors cannot completely solve the problem, but they represent an important tool to fight disinformation,”

Social media giants are starting to develop verification technologies to fight misinformation on their platforms. So similar detection methods to Assembler could become part of Facebook or Google in the coming years.