Deep Fake technology based consumer products become popular lately. They are currently being used mainly for entertainment purposes in mass market applications. For example, through deep fake web and mobile applications users can replace actor’s face with their own face in a favourite movie, animated gif or a photo.
These technologies however pose certain security threats. Thanks to deepfake technology it becomes possible for literally anyone to create an AI powered virtual impostor copy of a real person. The AI generated photos, videos and voice become virtually indistinguishable from those of a real person thanks to deep fake technology.
Deep fake technologies pave the way for multiple potential vulnerabilities in automated and human assisted ID verification, for example, as using real-time deepfake software one can now call another person and easily impose as someone they know and trust. Remember in Terminator movie when one of the terminators has changed their voice when responding to the phone call? This technology is now widely accessible. We can only predict its use will grow in attempts to gain unlawful access, social engineering and any other criminal and/or socially unacceptable purposes.
Deep fake technologies may also be used as a tool for fake news and provocations for example to swing public opinion during elections etc. Some of you would have seen the deep fake demo of President Obama which looks very realistic. In fact what happens there is another person’s voice recording has been transformed by an artificial neural network (deep leaning model) producing the result indistinguishable from the real Barack Obama video recording including face features, movements, mimics and even his voice.
DeepX team possesses expert knowledge in Deep Fake technology and we have consulted some early pioneers of this technology in the entertainment sector. While our team and our platform are not focusing our interests on this area, we offer the solution of deep fake detection.
By the nature of how human brain works, it successfully attempts to cover up imperfections of human eye, lighting conditions, lack of data and other natural limitations. As result, human brain actually helps the deep fake products to work, “forgiving” them certain imperfections.
Unlike human brain, machine learning models detect the imperfections of deep fake images and videos which allows DeepX to offer the capability of automated deep fake detection. The combined power of computer vision and audio recognition technologies together with DeepX team experience and knowledge of deepfake “kitchen” we are offering a solution that works in real-time with great accuracy and will protect the customers against the threats of deep fake technology.
