CACI: Make Your Own and Defend Against Deepfake Video/Images

CACI: Make Your Own and Defend Against Deepfake Video/Images

Presented by Andrew Mendez.

New cyber threats, such as DeepFakes, use machine learning to manipulate images and video recordings and create misleading content. Existing DeepFake tools enable the manipulation of recorded presentations of politicians or celebrities and generate realistic video and audio recordings by simply altering the transcript. These tools use state-of-the-art Deep Neural Networks, which can make their fabrications imperceptible to the human eye. DeepFakes pose significant threats to online social media ecosystems by creating greater distrust towards media organizations and enabling new kinds of large-scale information operation campaigns.

The purpose of this presentation is to: 1. Provide an overview of the progression of research in the field of DeepFakes; 2. Discuss how DeepFakes are being used on social media platforms, and 3. Describe how to defend against this threat. Additionally, a hands-on demo will be provided to engage the audience in how to create a simple DeepFake using Generative Adversarial Networks (GANs) and detect DeepFakes using Convolutional Neural Networks (CNNs).