Artificial Intelligence

How Tech Can Help Solve the Problem of Deepfakes

Overview Synthetic media can spread disinformation and even influence history records in the future. But technology can help detect deepfakes and protect real information.

Hyper-realistic artificial intelligence (AI)-generated videos can be created to show someone saying or doing something they never said or did. Deepfakes, as they are called, may soon be advanced enough to sow seeds of doubt in audiences’ minds. Can people trust what they are seeing on television if everything — information that can reflect both negatively or positively on someone — could be created artificially?

Rapid advancement of synthetic media could potentially allow creators to forge fake history records. Especially with the word’s growing digital focus, online archives may face troubles because of deepfakes.

But technology can provide solutions to the challenge of disinformation posed by deepfakes.

Video Authentication

CES® exhibitor and Consumer Technology Association (CTA)® member Microsoft developed the Microsoft Video Authenticator, which can analyze a still photo or a video and provide a percentage chance that the media is artificially manipulated or created.

The technology works by detecting the boundary of the deepfake creation and the blended original elements perhaps unseen by the naked human eye.

Microsoft’s authentication technology also includes tools that enable content producers to add digital hashes or certificates to a piece of content, which is then checked by a reader extension that confirms the certificates to check the authenticity of the content.

Biometric Signature Forensics

A digital forensics technique that harnesses machine learning to analyze a specific individual’s style of speech and unique body movements can create a biometric signature for that individual. This is especially useful for politicians and celebrities.

Because deepfakes cannot model all the complex and subtle signals of a specific person, mostly relying on face swaps, the technique can identify when speech or body movements are faked by digital puppets.

Blockchain Timestamps

The immutable quality of blockchain technologies allows us to establish an unchangeable timeline of digital events. Through an encrypted distributed ledger, consumers are able to determine if and how digital content has been modified.

Blockchain timestamps can essentially assign a digital fingerprint to every piece of digital media when it is uploaded and distribute it across the internet. Initial cryptographic hashes, and their following hashes should the file be modified, would allow users to verify the authenticity of the content they are viewing.

As deepfakes technology becomes more advanced — even being used in the entertainment world, perhaps — researchers and users will also define and improve the ways to protect digital records of history that people are able to trust.

You May Also Be Interested In ...

In Knowledge Graphs We Trust: Artificial Intelligence 2.0

Read more arrow-black

Did We Break Machine Learning?

Read more arrow-black