As a result of continuing advancements in neural networks, deep fake media has become increasingly convincing and easy to produce. Experts have warned of the impact this could have on elections and personal security. Additionally, deepfakes also pose very real threats to businesses and global markets, although these threats receive far less attention. Hacker and Security evangelist Alyssa Miller will analyze the technology behind creating deep fake media, showing how Generative Adversarial Networks (GAN) create convincing fake videos and audio from very limited samples. She will examine research into both low-tech and AI/ML based detection methods and counter measures, including leveraging the same neural network approaches being used to create deep fakes to help detect them. She’ll continue by discussing the theory and research behind
countermeasures such as Adversarial Perturbations and show how they can defeat facial recognition algorithms that deepfake generation relies on. Finally, Alyssa will present methods being developed to help certify the authenticity of real media.
As she concludes, Alyssa will offer up a hopeful viewpoint of the good that can be accomplished through the use of deepfake technology. From its use in entertainment, to improved analysis of medical imaging and even how GANs are being leveraged in malware identification.