Pushing the Frontiers of DeepFake Detection with DL and Explainability
Pushing the Frontiers of DeepFake Detection with Deep Learning and Explainability with Dr. Sherin Mathews, Senior Data Scientist at McAfee
Join us on Wednesday, Jan 5th, 6:00 – 7:00 pm PST to learn about the future of DeepFake detection frameworks from Dr. Sherin Mathews, Senior Data Scientist at McAfee.
Deepfake refers to fake content that is manipulated to look authentic using a technology called Generative Adversarial Networks. DeepFake Technology does have the potential to contribute to uncertainty and distrust of media content thereby posing a significant challenge to democracy. Credible yet fraudulent audio, video, and text will have a much larger impact that can be used to ruin celebrity and brand reputations as well as influence political opinion with terrifying implications. Imagine a dark web economy where DeepFakers produce misleading content that can be released publicly to influence major decisions. DeepFakes might touch all areas of our lives, and even basic protection is essential going forward.
This talk presents robust computer vision and deep learning DeepFake detection frameworks to exploit hidden patterns and manipulated parts that play a key role in authenticating original media files. To ensure the prediction results of the deep learning framework & the origin of solutions for each prediction are interpretable, a model-agnostic explainability framework is employed. Explainable Artificial Intelligence (XAI) creates explainable models while maintaining a high level of learning performance, thereby enabling users to understand and appropriately trust the underlying models. Having explanations for each prediction helps us better understand the model as it allows visualization of the layers and filters of the networks. Thus, explainable models provide insights into deep learning models thereby improving trust in the DeepFake predictions.
More on Dr. Sherin Mathews:
Dr Sherin Mathews is a Senior Data Scientist within the Office of the CTO for Intel Security/ McAfee. In this role, she creates and develops new machine learning/deep learning models to improve and increase the effectiveness of cybersecurity products. Her work involves research, implementation, and evaluation of next-generation security solutions with moonshot-style innovations, and potential subsequent integration of successfully POC’ed features into the product.