New chip technology aims to expose deepfakes at source

Could cryptographic signatures be the end of deepfakes?

A prototype microchip offers a potential defence against the growing threat of deepfakes.

Scientists at ETH Zurich say the technology, designed to verify the authenticity of images, videos and audio at the point of capture, could help restore trust in digital content, as AI continues to get better at creating convincing forgeries.

AI-generated media, often referred to as deepfakes, has become increasingly sophisticated in recent years.

Fabricated political speeches, altered footage from conflict zones and misleading social media posts are widespread online. Experts warn that such content is not only deceiving audiences but fostering a broader scepticism towards genuine information.

The newly developed chip aims to tackle this problem by embedding a cryptographic signature directly into data at the moment it is recorded.

This means that every image, video or audio clip captured using the sensor carries a unique digital marker verifying its origin and integrity.

"If data is signed the moment it is captured, any later manipulation leaves traces," said Fernando Cardes, one of the researchers behind the project.

The system works by integrating security functions within the sensor itself. This allows it to generate a signature that confirms the source of the data, the time of capture and whether any changes have been made afterwards.

According to the researchers, altering such data without detection would require physically tampering with the chip, a complex and resource-intensive task (although we will note that criminals don’t lack resources or expertise – Ed.).

"To manipulate the data, the chip would have to be physically attacked, requiring a massive technological effort," Cardes said, adding that such efforts would make "mass generation of manipulated content for social media platforms practically impossible."

The team suggests these digital signatures could be stored in a publicly accessible, tamper-proof database, such as a blockchain. Users could then verify the authenticity of content by comparing it with the stored record.

Felix Franke, who co-developed the technology and now works at the University of Basel, said the approach reduces reliance on trust in intermediaries.

"Trust in digital content is eroding," he said.

"We wanted to create a technology that gives people a way to verify whether something is genuine."

The chip could, in principle, be integrated into a wide range of devices, including smartphones, cameras and other sensors.

Social media platforms might automatically check uploaded content, flagging material that lacks verified signatures. Although that would require them to care less about engagement.

Where automated systems are not in place, journalists, researchers and public authorities could use verification tools to assess authenticity independently.

The prototype chip demonstrates that the concept is technically feasible. However, further development will be required before it can be deployed commercially.

The researchers say they are confident the technology can be brought to market using existing manufacturing processes and have already filed a patent application.

The announcement comes amid growing international concern about deepfakes.

Last month, the UK's data protection regulator, the Information Commissioner's Office, warned tech firms about the rise of AI-generated images of real people without consent, stressing that such content falls under data protection laws.

Meanwhile, the UK government has launched a partnership with Microsoft, academic institutions and industry experts to develop a national system for detecting manipulated media online.

Officials say the initiative will establish common standards to identify deepfake content.