If you watch a lot of entertainment and popular media, chances are you may have heard about deep fake technology. Videos where the face of an actor or celebrity is replaced with another to make it so convincing that one would not be able to tell the original footage had been tampered with. Such videos gain massive traction on YouTube with an example being of a store robbery stopped by the iconic Keanu Reeves. The video grew viral for the actor’s heroic actions until it was revealed it was not really him but in fact, a lookalike and impersonator whose face was accurately ‘deepfaked’.

Deepfake explained

Deepfake technology works with the help of AI to superimpose images of one party onto video footage of another. Machine learning algorithms are trained to help accurately map and create surprising new manipulated videos using techniques such as generative adversarial networks. What makes these videos amazing due to being unnoticeable to the untrained eye is also what makes it a potential cybersecurity threat especially with the technology being open source and available to virtually anyone.

The research found that more than 95% of 15,000 videos had misused Deepfake technology to replace faces of celebrities onto adult actors. The technology also had been involved in political scandals were fake footage of german chief-executive was used to fraudulently demand nearly $250,000 from the head of a UK-based energy plant. Such alarming crimes demand effective security measures that help catch tampered footage before serious damage is done.

Enter Online Facial Verification

Facial verification, a form of biometric authentication, helps to identify individuals using their facial profiles. The AI-powered software detects facial features with the help of liveness detection and depth-sensing. Although its implications are intended for the sake of security, its capabilities allow for it to be used in other kinds of applications.

Facial verification solutions all work in a similar manner which can be explained in the following steps:

- Samples of the face are first captured for a set duration - Facial features are extracted and made a template based on the collected information - The newly made template is cross-examined with other recorded data - The template is then used to determine the user’s face in newly recorded samples

How Facial Verification Solutions aim to stop deepfake attacks

Technology makes it harder for a person to believe what their eyes perceive as real anymore. For the sake of entertainment, such technology may amaze people but when used wrongly it may imply huge consequences. Potentially, deepfake may be used to bypass facial recognition software the more accurate the deepfake is. This puts a lot of companies and people’s lives and possessions at risk.

However, deepfakes may not be so effective when used in live footage as it requires immense hardware power to accurately make fake footage in real-time. This is where Online Facial Verification comes in handy to tackle such security risks as AI-powered facial recognition uses features such as liveness detection and depth-sensing to allow for more security where regular 2D facial verification may be fooled due to Deepfake’s capabilities.

Many giants such as Facebook are investing in effective facial verification and deepfake detection projects for the sake of the company and users’ security. While on the other hand, deepfake technology, too, receives research and development even though it may be for the sake to make cutting edge entertainment such as making realistic posthumous scenes of long-dead actors in upcoming Hollywood movies. But this quest for advancement only raises the possibility of threat due to malicious usage which only makes the need for online facial verification even more crucial, making it then a never-ending battle.

Share this project: