Overview
Save Big on Coursera Plus. 7,000+ courses at $160 off. Limited Time Only!
Explore a conference talk that delves into the vulnerability of neural networks to adversarial examples and critically examines the effectiveness of defensive distillation. Learn about three new attack algorithms that successfully generate adversarial examples for both distilled and undistilled neural networks with 100% probability. Discover how these attacks are tailored to different distance metrics and compare their effectiveness to previous adversarial example generation algorithms. Gain insights into the proposed use of high-confidence adversarial examples in a transferability test that can potentially break defensive distillation. Understand the importance of this research in establishing benchmarks for future defense attempts aimed at creating neural networks resistant to adversarial examples.
Syllabus
Towards Evaluating the Robustness of Neural Networks
Taught by
IEEE Symposium on Security and Privacy