Adversarial Attacks in Machine Learning and How to Defend Against Them

Notes from the Keynote Speech by Professor Ling Liu at the 2019 IEEE Big Data Conference

Big Data powered machine learning and deep learning has yielded impressive advances in many fields. One example is the release of ImageNet consisting of more than 15 million labelled high-resolution images of 22,000 categories which revolutionized the field of computer vision. State-of-the-art models have already achieved a 98% top-five accuracy on the ImageNet…