How to make sure you’re talking to a real person online? IDEAS NCBR researchers have developed a prototype solution

In recent years, we have been hearing more and more about cases of fraud using deepfakes. In Poland, in March 2024, fake videos with the participation of InPost CEO Rafał Brzoska, Andrzej Duda and Robert Lewandowski promoted fraudulent “investment programs” intended to extort money. In January, a bank employee in Hong Kong withdrew $25 million to fraudsters who impersonated the company’s chief financial officer during a video conference. Deepfakes are becoming more advanced and more difficult to detect, presenting us with new digital security challenges.

In response to these threats, scientists from IDEAS NCBR created an experimental project called ProvenView. The creators of the solution are postdoc Shahriar Ebrahimi and PhD student Parisa Hassanizadeh, researchers from Iran working in the “System security and data privacy” research group led by Professor Stefan Dziembowski. During the ZK Hack Kraków hackathon in May 2024, ProvenView was awarded in the “best proof of provenance project” category.

‘A decade ago, Zero-Knowledge Proofs (ZKP) technology was considered a purely theoretical field, closer to pure mathematics than to the applications of computer science. In recent years, we have witnessed a dramatic change in this field. This happened thanks to many breakthrough improvements in this technology, as well as thanks to its popularization thanks to blockchain. The ProvenView solution is yet another spectacular example of how ZKP can be used in the fight against the threats of the modern digital world,’ says Professor Stefan Dziembowski, leader of the “System security and data privacy” research group at IDEAS NCBR.

Checking the authenticity of a video at your fingertips

ProvenView can be used both in everyday online interactions and in protecting creators’ content. For example, after an online meeting ends, the user will be able to check whether the video of his interlocutor actually came from his webcam. But this is not the end of its potential applications.

‘This solution may be useful both for individual users who could use it to authorize online meetings, and for content creators who could thus strengthen the protection of their image published in a YouTube video against theft and use, e.g., in pornography,’ says Shahriar Ebrahimi, postdoc at IDEAS NCBR.

Authorization without an intermediary

ProvenView is a prototype software that uses Zero-Knowledge Proofs (ZKP) for authorization without having to trust an intermediary. What exactly is ZKP? It is a cryptographic technique that allows you to prove the truthfulness of certain information without revealing any details about it.

Here’s a simplified example of how it works: In the first step, the author of the original video creates digital proof of its authenticity. It’s a digital record that says, “This video is real and I created it.” The software enabling such recording may be built into the video recording application or be a separate program. Counterfeiting proof of authenticity based on ZKP is extremely difficult due to the advanced cryptographic mechanisms used in them. In the second step, when someone would like to check the authenticity of the video, we wouldn’t have to show them the entire video. Instead, we would show them digital evidence. And this is where Zero-Knowledge Proofs helps us: thanks to advanced mathematics, this proof can be verified without the need to watch the video itself.

‘The advantage of ZKP (Zero-Knowledge Proofs) is that the person creating the proof of authenticity does not need to have external trust confirmations, such as when using an electronic signature. Anyone can generate such proofs of authenticity on their own device. This is called “client-side proof generation”. Thanks to this, the solution is very flexible and can serve many people at the same time, without burdening the network. Regardless of whether evidence is generated by 10 people, 10 thousand or even 10 billion people, the network load does not change,’ explains Shahriar Ebrahimi, postdoc at IDEAS NCBR. ‘In the future, our solution could become an add-on for web browsers of video editor software.’

Further development of the project

ProvenView is an experimental solution and not yet ready for implementation. The project at its current stage is a kind of proof-of-concept that such methods can be used in practice.

’Today, proofs of authenticity can only be generated once video creation has been completed. Additionally, proving authenticity is very time-consuming and cannot be done online in real time. With the technology currently available, we can most quickly prove the authenticity of a low frames per second (fps) video. Let’s assume the original meeting video was recorded at 30 fps. First, we very quickly check to see if there is any abnormal activity in the 2 fps grayscale version. Later, for greater certainty, we check the authenticity of the color video. Ultimately, the proof would be available, for example, an hour after the conversation,’ adds Shahriar Ebrahimi, postdoc at IDEAS NCBR.

Thanks to further work on efficient recursive proofs, scientists from IDEAS NCBR hope that in the future it will be possible to prove video authenticity in real time.