The tool, accessible via the Surf browser or as an extension, has the ability to detect AI deepfakes with up to 98% accuracy. Amazing, right? And how would you use it?
TL;DR
- Surf Security has implemented a tool capable of detecting deepfakes with 98% accuracy.
- The tool detects AI-generated imitations across different languages and accents by analyzing audio frames.
- Surf Security plans to offer this tool in full version early next year.
A revolutionary tool against deepfakes
In a world where artificial intelligence permeates every aspect of our lives, the fantasy of a cyberspace without deepfakes is becoming a reality. And it is to the British company Surf Security, a new cybersecurity startup based in London, that we owe this technological progress.
The company Surf Security developed a browser; the first in the world, they claim, with an integrated function designed to detect deepfakes generated by Artificial Intelligence (AI). This technology, militarized in their terms, uses what we call “state spatial models”to detect AI-generated clones across different languages and accents. It analyzes audio frames for inconsistencies.
As Surf Security CTO Ziv Yankowitz explains, the neural network training uses deepfakes created by leading AI voice cloning platforms. For him, the real strength of their system is its speed and precision. He claims that he is “capable of spotting a fake audio track in less than two seconds”.
Put an end to deepfakes
It should be noted that the software is capable of identifying whether the audio – whether recorded or live – is provided by a human or an AI. Separately, the company says it is also adding an AI-generated image detector to its browser in the future.
Deepfakes, which use AI to create falsified audio or video, continue to pose a growing danger. They have been used to commit large-scale fraud, incite political unrest through fake news, or destroy reputations through the creation of false or harmful content.
Surf Security launched this deepfake detector with the aim of protecting businesses, media organizations, police and military around the world from this growing risk of AI cloning.
This is a big challenge in the face of AI technology that is improving every day, admits Ziv Yankowitz. However, he promises that the full version of their deepfake detector will be available next year.
